A Sociophonological Analysis of Mersea Island Englishroa.rutgers.edu/content/article/files/1208_amos_1.pdf · IV ABSTRACT This thesis presents a socio-phonological analysis of three

A Sociophonological Analysis

of Mersea Island English:

An investigation of the diphthongs

(), () and ()

Jennifer Amos

A thesis submitted for the degree of PhD Linguistics

Department of Language and Linguistics

University of Essex

September 2011

II

ACKNOWLEDGEMENTS

Firstly, I would like to extend my thanks to all the people of Mersea Island who willingly

gave up their time to help with this study. They welcomed me into their homes, gave me

endless cups of tea and provided me with any information I needed. Also, to all those at

the Mersea Museum, thank you for granting me access to the archive recordings and for

being so patient when I needed help. This research could not have been completed

without you all.

I also wish to thank to the E.S.R.C. for their financial support (Award Number PTA-031-

2006-00256) and my supervisors, Prof. Dr. David Britain and Dr Wyn Johnson, for their

continuous encouragement and faith in my research.

To David, my family and my friends, what can I say? Thank you for your love and

tolerance, for sharing the laughter and the tears, and for always being there.

Finally, to my dad, Adrian – Thank you for being the most hilarious proof-reader anyone

could wish for!

- J.A.

III

“Is the bottle half-full or half-empty? Neither. There is always something in place of

that space where nothing can be seen, whether it be air, or something other. Therefore the

bottle is always full, but, like life, it is not always full of what you expect.”

- Michelle-Lee Phelan

IV

ABSTRACT

This thesis presents a socio-phonological analysis of three diphthongs in Mersea Island

English (MIE). Mersea Island is situated off the North East coast of Essex in South East

England. Socio-economically, the Island has seen dramatic change over the past century.

Originally quite isolated and economically largely self-sufficient, social and demographic

changes have led to significantly greater contact (both social and linguistic) with the

mainland. Thus, parallels may be drawn between the developments of Mersea Island and

those of, for example, Martha’s Vineyard and Ocracoke Island in North America.

The nature of these social changes will be evaluated alongside the analysis of the

diphthongs MOUTH, PRICE and CHOICE. As a result, the direction of change and typologies

of variation present in the speech of three generations of Mersea Islanders will be

established, and the relationship between the dialectological findings and external

sociolinguistic factors will be explored.

The results from the sociolinguistic data analysis will then be considered in light of

naturalness and phonological theory. Using an approach combining the mechanics of

Optimality Theory (OT) and the principles of Dispersion Theory, a three-tiered model

will be constructed which effectively represents the interface between internal (linguistic)

factors and external (sociolinguistic) influences on language variation and change. It will

be seen, through the application of this model to MIE that, while some variation may be

generated at the linguistic level, other variation may be attributed to selections based upon

sociolinguistic considerations. In addition, the incorporation of variation at each level of

the model allows for a distinction to be made between language change, which is

V

generated through linguistic motivations, and change which is motivated by external

sociolinguistic desires.

VI

TABLE OF CONTENTS

Abstract IV

Introduction 1

Chapter 1 11

Mersea Island – The Community: it’s social history and

demographic change

1.1 The Island 12

1.2 Access 15

1.3 The Island’s Geographic Development 15

1.4 Modern Mersea 18

1.5 Summary 29

Chapter 2 30

Methodology

2.1 Data Collection 32

2.1.1 The Interviews 32

2.1.2 The Informants – General Overview 35

2.1.2.1 The Museum Speakers 37

2.1.2.2 The Older Generation 38

2.1.2.3 The Younger Generation 38

2.2 Social Factors 39

2.2.1 Class 39

2.2.2 Age and Gender 41

2.3 The Variables – Extraction, Analysis and Classification 42

2.3.1 Linguistic Constraints 42

2.3.1.1 Syllable Type 43

2.3.1.2 Phonological Environment 44

2.3.1.2.1 The Distribution of /a/ 44

2.3.1.2.2 The Distribution of /a/ 45

2.3.1.2.3 The Distribution of // 46

2.3.1.3 Stress 46

2.3.1.4 LIKE 47

2.4 Historical Data 47

2.4.1 Ellis (1889) 48

2.4.2 Wright (1968) 49

2.4.3 Kurath and Lowman (1970) 50

2.4.4 The Survey of English Dialects (SED) 51

2.5 Summary 52

VII

Chapter 3 53

The Historical Derivation and Variation of the Variable ()

3.1 Historical Sources 54

3.2 Variation of (a) 60

3.2.1 Historical Variation 64

3.2.1.1 Ellis (1889) 64

3.2.1.2 Wright (1968) 64

3.2.1.3 Kurath and Lowman (1970) 65

3.2.1.4 The Survey of English Dialects (SED) 65

3.2.1.5 Summary 66

3.2.2 Modern Variation Studies – The South East 67

3.2.2.1 The East Anglia Fens 67

3.2.2.2 Colchester 69

3.2.2.3 Basildon 70

3.2.2.4 Summary 71

3.2.3 Contemporary Studies of Insular Varieties 72

3.2.3.1 Martha’s Vineyard 72

3.2.3.2 Ocracoke and Smith Island 74

3.2.3.3 The Falkland Islands 76

3.2.3.4 St Helena 78

3.2.3.5 Tristan da Cunha 80

3.3 Summary 81

Chapter 4 83

4.1 () - An Analysis of Social and Linguistic Factors 84


4.2.1 Age 86


4.3 Linguistic Factors 92

4.3.1 Syllable Type 92

4.3.2 Preceding Phonological Environment 94

4.3.3 Following Phonological Environment 98

4.3.3.1 Following Voice 100

4.4 Summary 104

Chapter 5 106

The Historical Derivation and Variation of the Variable (a)


5.2 Variation of (a) 111

VIII

5.3 Historical Variation 114

5.3.1 Ellis (1889) 114

5.3.2 Wright (1968) 115



5.3.5 Summary 117

5.4 Some More Recent Studies 117

5.4.1 Norwich 118

5.4.2 The East Anglian Fens 120

5.4.3 Other South-Eastern Variation 121

5.4.4 Summary 123

5.5 Contemporary Studies of Insular Varieties 124

5.5.1 Martha’s Vineyard 124

5.5.2 Ocracoke and Smith Island 126

5.5.3 The Falkland Islands 128

5.5.4 St Helena 129

5.5.5 Tristan da Cunha 131

5.6 The Special Status of LIKE – Variation in New Zealand

English

132

5.7 Summary 133

Chapter 6 135



6.2.1 Age 138




6.3.2 Preceding Phonological Environment 141

6.3.3 Following Environments 142

6.3.3.1 Voice 143

6.3.3.2 Manner 146

6.3.3.3 Following Place of Articulation 146

6.3.3.4 Following /l/ and the Status of ‘Island’ 147

6.3.4 LIKE 152

6.4 Summary 155

IX

Chapter 7 157

The Historical Derivation and Variation of the Variable ()


7.1.1 // and //, a Temporary Merger? 159

7.2 Historical Variation 162

7.2.1 Ellis (1889) 162



7.3 Modern Variation Studies 164

7.3.1 British English 165

7.3.2 International Insular Variation 166

7.3.2.1 The Falkland Islands 166

7.3.2.2 St Helena 166

7.3.2.3 Tristan da Cunha 167

7.4 Summary 167

Chapter 8 169



8.2.1 Age 170




8.3.2 Phonological Environments 176

8.3.2.1 Preceding Place of Articulation 176

8.3.2.2 Preceding Manner of Articulation 179

8.3.3 The BOY Effect 180

8.4 Summary 183

Chapter 9 185

MIE - Sociolinguistic Considerations

9.1 The Direction of Change 186

9.2 Performance and Enregisterment 190

9.2.1 Performance, Enregisterment and Mersea Island English 193

9.3 Contact, Diffusion and Supra-localisation 200

X

Chapter 10 209

Socio-Phonology: the construction of an integrated model

10.1 What is Optimality Theory? 211

10.1.1 The Nature of Constraints 218

10.1.1.1 Faithfulness Constraints 218

10.1.1.2 Markedness Constraints 221

10.1.2 Optimality Theory and Language Variation 222

10.1.2.1 Tied Violations 223

10.1.2.2 Multiple Grammars 224

10.1.2.3 Stochastic OT 226

10.2 Dispersion Theory 229

10.2.1 Dispersion Constraints 231

10.3 A Three-Part Phonology 236

10.3.1 The Inventory 237

10.3.2 The Realisation 239

10.3.3 The Evaluation of Surface Constraints 241

10.3.4 Evaluation of this Approach 243

10.4 An Inventory- Based Model of Mersea Island English 254

10.4.1 The Inventory 254

10.4.2 The Realisation of Contrasts 254

10.4.2.1 Canadian Raising 255

10.4.2.2 The Dorsal Effect - A lowering of diphthong nuclei

after /k/

257

10.4.2.3 Smoothing - // to /:/ before coda /l/ 259

10.4.2.3.1 Smoothing - Proposal 1 260

10.4.2.3.2 Smoothing - Proposal 2 262

10.4.2.4 Island Monophthongisation 263

10.4.2.5 BOY 266

10.4.3 The Older Speakers – multiple grammars or variable

ranking?

270

10.4.4 Evaluation of Surface Constraint (ESC) 272

10.5 The Sociolinguistic Level of Speech Production 275

10.5.1 Introduction 275

10.5.2 Communicative Competence 276

10.5.3 Where does that leave phonology? Information through

exemplars

280

10.5.4 How this can be implemented through the three-tiered

model

284

10.5.5 The Activation of the Sociolinguistics Matrix and

Evaluative Parameters

291

10.6 The Representation of Types of Change 294

XI

Conclusion 304

Bibliography 308

Appendix A 319

The complete numerical data sets pertaining to the analysis of

the variable ()

Appendix B 339

The complete numerical data sets pertaining to the analysis of

the variable ()

Appendix C The complete numerical data sets pertaining to the analysis of

the variable ()

358

Appendix D

Examples from the Mersea data for each of the variables

examined with respect to preceding and following environments

382

Appendix E

The (stressed) vowel inventories of two Mersea Island speakers

388

XII

LIST OF MAPS, FIGURES, GRAPHS AND TABLES

Maps

1.1 a Location of Mersea Island on the South Eastern coast of England 13

1.1 b Location of Essex in relation to London and surrounding counties 13

1.1 c Location of Mersea Island in relation to the Essex coast 14

1.1 d Close up of Mersea Island 14

2.1 Location of Mersea Island, West Bergholt and Titpree 51

Figures

3.1 The developmental path of /a/ following the lower-centralise-

lower route

55

3.2 The developmental path of /a/ following the centralise-lower-

lower route

55

3.3 The Stages in the Development of Middle English u 57

3.4 The path of the diphthong shift in both front and back vowel

systems

58

3.5 The alternative development of () 59

5.1 The developmental path of /a/ following the lower-centralise-lower

route

108

5.2 The developmental path of /a/ following the centralise-lower-lower

route

108

5.3 The stages in Development of Middle English 109

5.4 The possible developmental path of /a/ in the Essex dialect 111

7.1 The development of /a/ and // demonstrating the potential overlap

of phonemes

160

9.1 The direction of change for the nuclei of Mersea Island //, // and // diphthongs

187

9.2 An S-Curve: pattern of change for the variable (wh) in Central

Canada

189

10.13 A representation of vowel space according to Flemming (2004) 232

10.15 A representation of vowel space according to Flemming (2004) 238

10.19a Illustration of Standard English Diphthongs 246

10.19b Illustration of Standard English centering diphthongs 246

10.20 A simplified segmentation of vowel space used for the analysis of

MIE

247

10.38 A comparison of underlying Grammars for the Older Mersea

generation

271

XIII

Graphs

1.2 The population of Mersea Island 18

1.3c The 2001 population breakdown of West Mersea, Colchester and

the East of England

20

1.4b The net population change in Islander numbers aged 15-24

between July 2001 and June 2009

21

1.5b The net population change for Islanders aged 25-44 between July

2001 and June 2009

23

1.6c The Age of Businesses registered in 2010 in West Mersea,

Colchester and the East of England

25

1.8 Percentage of the West Mersea workforce population (aged 16-74)

according to distance travelled to work

28

4.1c Overall percentage of (a) variants by age 87

4.2c Overall percentage of (a) variant by age 88

4.3 Percentage of (a) variants according to age and gender 90

4.4 Overall percentage of (a) variants according to syllable type 92

4.5 Overall percentage of full and short diphthongal offglides according

to syllable type

93

4.6 Percentage of full and short diphthongal offglides according to age

and syllable type

94

4.7 Overall percentage of (a) variants according to preceding manner

of articulation

95

4.8 Overall percentage of (a) variants according to preceding place of

articulation

96

4.9 Overall percentage of (a) variants according to following manner

of articulation

98

4.10 Overall percentage of (a) variants according to following place of

articulation

100

4.11 Overall percentage of (a) variants according to the voicing of the

following segment

101

4.12 Percentage of (a) variants according to following voice by

Museum speakers

103

4.13 Percentage of (a) variants according to following voice by Older

speakers

103


Younger speakers

103

5.5 Variable () by class and style 119

5.6 Centralisation of (ai) and (au) plotted separately, according to year

of birth.

125

6.1b Percentage of (a) variants according to speaker age 138

6.2 Percentage of (a) variants according to age and gender 139

6.3 Overall percentage of (a) variants according to syllable type 140

6.4 Percentage of (a) variants according to syllable type and age group 141

XIV

6.5 Overall percentage of (a) variant according to preceding manner of

articulation

141

6.6 Overall percentage of (a) variant according to preceding place of

articulation

142

6.7 Overall percentage of (a) variants according to the voicing of the

following segment

144

6.8 Percentage of (a) variants according to following voice by Museum

speakers

145

6.9 Percentage of (a) variants according to following voice by Older

speakers

145


Younger speakers

145

6.11 Overall percentage of (a) variants according to following manner

of articulation

146

6.12 Overall percentage of (a) variants according to following place of

articulation

147

6.13 Overall percentage of (a) variants preceding /l/ 148

6.14 A comparison of the overall percentages of (a) variants according

to /l/ type

149

6.15 Overall percentage of Diphthongal vs. Monophthongal variants for

the token Island

149

6.16 Percentage of (a) variants for Island by Museum speakers 151

6.17 Percentage of (a) variants for Island by Older speakers 151

6.18 Percentage of (a) variants for Island by Younger speakers 151

6.20 An overall comparison of (a) variants according to the functions

of LIKE

154

6.21 A comparison of (a) variants according to the functions of LIKE

by Older speakers

155

6.22 A comparison of (a) variants according to the functions of LIKE

by Younger speakers

155

8.1b Percentage of () variants according to age 171

8.2 Percentage of () variants according to age and gender 172

8.3 Overall usage of () variants according to syllable type 173

8.4 Percentage of () variants according to age and syllable type 175

8.5a Percentage of () variants used by Mersea males according to age

and syllable type

175

8.5b Percentage of () variants used by Mersea females according to

age and syllable type

175

8.6 Overall percentages for () variants by Mersea Island speakers

according to preceding place of articulation

177

8.7 Percentage of () variants by Museum speakers according to

preceding place of articulation

178

8.8 Percentage of () variants by Older speakers according to


178

8.9 Percentage of () variants by Younger speakers according to


178

8.10 Overall percentages of () variants by Mersea Island speakers

according to preceding manner of articulation

179

XV

8.11 Percentage of () variants across lexically derived BOY tokens by

Museum speakers

181


Older speakers

181


Younger speakers

181

8.14 A comparison of () variants between BOY tokens and all other

tokens by Museum speakers

182

8.15 A comparison of () variants between BOY tokens and all other

tokens by Older speakers

182

9.3 The population of Mersea Island, based on Census data from 1801

to 2001

203

9.4b Percentage of the West Mersea workforce population (aged 16-74)


205

XVI

Tables

1.3a The 2001 population breakdown of West Mersea, Colchester and

the East of England.

19

1.3b The 2001 population percentages of West Mersea, Colchester and

the East of England.

19

1.4a The Net Migration Statistics, from the 2001 Census, for the 15-24

year old age group in the West Mersea

21

1.5a The Net Migration Statistics, from the 2001 Census, for the 25-44

year old age group

22

1.6a/b Data Demonstrating the Age of Businesses registered in 2010

with a comparison between West Mersea Colchester and the East

of England

24

1.7a Data representing types of employment for residents of West

Mersea, Colchester and the East of England

26

1.7b Percentages of types of employment, ranked according to the

percentages of Mersea residents

27

3.6 Survey of English Dialects variation of // when followed by // in

East Mersea, Tiptree and West Bregholt

66

4.1 a/b Token Numbers/ Percentages of () variants according to age

in Mersea Island English

86

4.2 a/b Token Numbers/ Percentages of () variants according to

diphthong type and age in Mersea Island English

88

6.1a Percentage of (a) variants according to diphthong type and age 138

6.19a/b Overall number/ percentage of (a) tokens for the token LIKE 153

8.1a Percentage of () variants according to age 171

9.4b Percentage of the West Mersea workforce population (aged 16-74)


205

1

Introduction

2

INTRODUCTION

“This place is even worse that you could imagine in your wildest dreams. No gas or

electric lights. From Sir J.H’s house to post office 2 ½ miles. Got down at 4 p.m. Have to

go to Colchester this afternoon, thank goodness”

From Jay (2005: article 60)

The above message was written on a postcard sent from East Mersea in 1914. It

demonstrates that, less than a century ago, there was a marked difference between city life

and country life in England. With no gas, electricity or even running water at this time,

together with only infrequent carrier carts as transport between the Island and the

mainland, the isolation of the Island community was quite marked. However, even

though the community was relatively isolated, the Island itself was a hive of activity and

the local oystering industry was developing an almost stereotypical association with

Mersea Island.

The following is an excerpt from a news article written in 1895 (see Jay (2005: article 80)

for the full text):

“...one can either hire a trap or go by carrier’s cart to a little place called West Mersea,

about ten miles distant, situate [sic] on the north side of Blackwater Creek, and an island

at high tide. The inhabitants are all occupied in Oyster Dredging”

Indeed, oystering is one of the Island’s oldest industries and the oyster shed, which is

referred to within the full version of the article cited above, still stands and is used by the

industry today. Despite this isolation, which sociolinguists such as Milroy (1987) claim

would have resulted in very strong dialect-reinforcing social networks, the Island has

gradually opened itself up over the last century. The early part of the twentieth century

3

witnessed an increase in public transportation and the occasional private motor vehicle for

more wealthy Islanders or visitors.

Throughout the second half of the twentieth century, significant changes had taken place

across the Island. Among these were the rebuilding of the Strood (the causeway road

which links the Island to the mainland) so that the Island was not cut off by the tides to

the same extent that it had been in the past, and the extensive construction of family

homes on the West side of the Island. The latter development resulted in a dramatic rise

in the population and, as a result of these cumulative factors, Mersea Island was no longer

such an isolated community.

In more recent times, West Mersea has been described by the Guardian in 2006 as:

“where the sea meets the sky and the sky meets the mud. It’s a remote, flat, watery end

zone of Essex”1

In addition, The Telegraph is among the papers which has promoted Island property to

their readers with thought provoking sub-headings such as:

“Is the Island of Mersea the perfect location for commuters in Essex?”2

and has described the road which links Mersea to the mainland as “the causeway to

heaven”3.

1 http://www.guardian.co.uk/lifeandstyle/2006/dec/16/foodanddrink2?INTCMP=SRCH –December 2006

2 http://www.telegraph.co.uk/property/3361460/Property-in-Essex-Causeway-to-heaven-a-16th-century-

house-with-Essex-appeal.html - May 2008 3 http://www.telegraph.co.uk/property/3361460/Property-in-Essex-Causeway-to-heaven-a-16th-century-

house-with-Essex-appeal.html - May 2008

4

This type of media coverage raises awareness of the Island for those outside the local area

and, as a result, the Island also plays host to a number of short term visitors. A number of

these visitors come to savour the produce of the local fishing and oyster industry. Indeed,

the Island houses what has become a nationally iconic seafood restaurant, The Company

Shed. This is run by a local family who have been working the Mersea waters for eight

generations4 and it is regularly featured in good food guides and newspapers, such as The

Guardian’s ‘Top Places to Eat by the Sea’.

It is this unique geographical setting and social and historical developments which make

Mersea Island an ideal community to study dialectologically. Dense multiplex social

networks (where network members share the same acquaintances and are linked to them

through a range of professional and social activities) are a characteristic of more isolated

communities. Through these types of networks, it has been shown that dialect features

and linguistic complexities are reinforced and, therefore, innovative variants are less

likely to permeate the dialect variety. However, once these networks begin to loosen and

their members establish contacts with those of other networks, linguistic innovations may

be introduced, accommodated and reinforced. Therefore, the rises in the residential

population, and other changes to Mersea’s socio-demographic and socio-economic

structure would lead us to expect a weakening of social network ties and a vulnerability to

innovative linguistic forms.

In situations of dialect contact, it has been observed that koineisation and levelling may

lead to the adoption of particular variants, such as those which are considered less marked

or those which represent a phonetically intermediate point between two or more input

4 Information from: www.thecompanyshed.co.uk/about - September 2011

5

variants. In addition, the process of regional dialect levelling has been shown to lead to a

reduction of differences between dialect varieties and, thus, the vernacular speech of a

region will gradually homogenise (Kerswill 2004:671).

However, each community will experience different types of contact. With respect to

insular communities, such as Martha’s Vineyard (Labov 1972, for example) and

Ocracoke Island (Wolfram 2008, for example) off the east coast of North America, a

significant level of contact comes from those short-term visitors to the Island and those

who permanently move to the Island from the mainland. The latter would particularly

result in long-term contact and accommodation. In these situations, investigations into

community affiliation and attitude have proved insightful. For example, correlations can

be found between strong local affiliations and a greater use of local or traditional variants.

Conversely, those who are more positive towards the incoming population (whether

visitors or permanent residents) are more likely to accommodate to the incoming variants

and adopt innovative forms.

However, the communities mentioned above each experience relatively high levels of

contact through tourism and mainland travel. Island communities such as St Helena and

Tristan de Cunha (Schreier 2010a; 2010b, respectively) represent more extreme isolation.

In these communities, the socio-cultural situations are relatively stable (though

historically this may not have always been the case) and, therefore, innovations generated

through high levels of contact with outsiders is less likely. As a result, we would expect

these dialect varieties to remain fairly stable following the initial settlements and

processes of contact and koineisation.

6

The investigation of these varieties, alongside that of Mersea Island English will enable

an investigation into whether the varying types of contact and population movements

have led to similar linguistic developments across these disparate insular communities,

particularly with respect to those whose insular status has been reduced in recent times.

Kerswill (2004) also notes that, with respect to koineisation and levelling processes, “the

degree of difference between the input varieties will affect the amount of accommodation

individuals will have to engage in” (2004:695). Thus, in a community such as Milton

Keynes, where the differences between the contact varieties are sub-phonetic, the

focussing towards particular variants is accelerated. This is likely to be the case with

respect to contact affecting the Mersea Island dialect, as the surrounding varieties contrast

only in linguistically subtle ways. Indeed, Trudgill (1986) notes that, in contact

situations, some forms are more likely to be maintained in place of others due to, for

example, their relative social saliency or linguistic markedness (where the latter tends to

be reduced during contact). For example, the quality of the MOUTH vowel in both Milton

Keynes and Reading has been shown to be moving towards a standard-like [au] form,

despite these communities having inputs of a [] ~ [] nature and contrasting social and

historical developments (Cheshire 1999). Taking into account data from communities

such as these, which contrast with respect to social history and the quality of input

variants during contact, we can assess how typologically similar the emerging dialects

are. Thus, if contact is seen to result in phonologically similar forms across dialects, we

may then be in a position to propose that the direction of change is being guided by

natural phonological desires which would otherwise be inhibited by social isolation.

7

The representation of natural phonological tendencies may be achieved through the

application of phonological theory to sociolinguistic data. For example, a framework

such as Dispersion Theory (which uses the basic mechanics of Optimality Theory)

presents the optimisation and naturalness of the underlying phonological system as its

central driving force. As a result, it aims to evaluate the well-formedness of the

phonemic system as opposed to evaluating individual sounds. Therefore, by considering

the system as a whole, we can look to explain why certain movements to underlying

phonemes may be preferred over others.

These social and linguistic considerations and observations lead to certain questions

which must be approached and considered in the sociophonological analysis of MIE:

1. If insular communities, by their very social structure, will act to reinforce and

stabilise linguistic features and complexities, what happens when contact with

speakers beyond the immediate community becomes established in the long

term?

2. When the strong original community structure begins to erode in favour of

wider regional allegiances, does the linguistic structure begin to move in the

direction of more supralocal dialects, or is it preserved by the community in

the form of identity markers?

3. Furthermore, can we simply account for any apparent change through

sociolinguistic factors relating to contact, local identity and attitudes, and

changes in regional affiliations?

4. Alternatively, in light of variation studies from a range of other disparate

insular communities that demonstrate similar outcomes from dialect contact,

can any changes be represented through phonological theory and the natural

motivations of the underlying linguistic system?

8

This thesis will aim to answer these questions among others in relation to Mersea Island

English, in particular the three variables (), () and (). In order to do this, it is

necessary to explore the historical developments that have taken place on the Island,

which may have influenced and shaped the community’s development over the years. In

addition to local history, the current social structure of the Island needs to be explored so

that the nature of modern Island life can be ascertained. These matters will be presented

and discussed in detail throughout Chapter 1.

Chapter 2 will focus upon the methodology employed in order to carry out the data

collection on Mersea. The nature of the sociolinguistic interviews will be highlighted, as

well as a detailed description of the informants used, and relevant social factors will be

discussed. In addition to these social considerations, the extraction and coding of the data

and the relevant linguistic constraints pertaining to the individual variables will be

examined. This will lead to the examination of each variable’s unique distributional

pattern within the linguistic system. Chapter 2 will close with a brief outline of the

historical dialect sources employed throughout the thesis, in order to provide backgrounds

to each variable’s history.

Chapters 3 and 4 will focus on the variable (). Chapter 3 will detail the historical

development of () as well as results from modern variation studies, in order to give an

idea of any behavioural patterns with which it is associated. These studies will focus,

primarily, upon data from East Anglia and the South East of England in order to set the

relevant dialectological scene in relation to Mersea Island. However, this chapter will

also include observations from international insular studies (in which British English had

9

an input). This is so that any similarities or differences in ()’s behaviour can be

observed and taken into account in other isolated social environments. The information

presented in this chapter will thus provide the basis for the analysis of () in MIE, which

is presented in Chapter 4. This analysis will include results pertaining to the social

factors of age and gender, as well as any applicable linguistic factors.

The structure of these two chapters is replicated in Chapters 5 and 6 and Chapters 7 and 8.

The former pair will discuss and analyse the variable (), while the latter pair will

concentrate on the variable ().

In light of the analyses presented in Chapters 4, 6 and 8, Chapter 9 will consider the

implications that the results have in light of sociolinguistic theory. The nature of

language and dialect change will be discussed, as well as how certain dialect features,

which are enregistered and adopted in dialect performances, may act as preservers of

traditional features in certain contexts. Included in this discussion will be the distinction

between the enregisterment of contextual (or specific) linguistic processes, general

linguistic processes and lexically-based processes and how these can be represented in the

results from data analysis. Finally, aspects of dialect contact, social mobility, and supra-

localisation will be discussed in light of the changes highlighted in the Mersea data.

This discussion will lead to the introduction of ‘naturalness’ as a motivation in linguistic

change, and this will be one of the central themes of Chapter 10. Following a detailed

introduction to the central principles of Optimality Theory and Dispersion Theory,

Chapter 10 will present the construction of an integrated phonological and sociolinguistic

model. This model, which truly represents the socio-phonological interface, will use data

10

patterns highlighted in the previous chapters to demonstrate its application not only to

sociolinguistic variation, but also dialect change. Finally, in light of this model, which

consists of three separate levels of processing, the mechanics of dialect change and

language change will be shown to apply to each level of processing, depending on the

type of change which is being modelled (that is, whether the change is instigated and

motivated internally or externally).

11

Chapter 1

Mersea Island – The community:

its social history and demographic

change

12

1. MERSEA ISLAND – THE COMMUNITY5

This chapter will discuss Mersea Island with respect to its history and socio-demographic

make-up and development. From its relative isolation in the late 19th

century to its more

extensive interconnection with the mainland in the present day, this chapter examines the

evidence for social and demographic change that provides the context for the linguistic

variation and change that I will investigate later.

1.1

The Island

Mersea Island is situated amongst the alluvial plains of North East Essex (in South East

England) between the estuaries of the River Colne and the River Blackwater. It is the

most easterly inhabited island of the British Isles and, at its highest point, sits only

twenty-one metres above sea-level. Its museum highlights evidence of Celtic inhabitation

as well as their salt panning and ceramic industries. In addition, Karbacz notes that,

although Mersea was under early Roman occupation, and is situated very close to the

Romans’ important urban centre Camulodunum (modern day Colchester), “we still call

Mersea by its Saxon name, Meres-ig, the island of the Mere” (1999:14)

5 Note: All the statistical data presented in this chapter are based on those held on the Office for National

Statistics web-site (www.neighbourhood.statistics.gov.uk/dissemination/) unless otherwise stated.

13

1.1 a) From: http://en.wikipedia.org/wiki/Mersea_Island

1.1 b) From: www.itravel.co.uk/maps/england

Mersea Island

The map in 1.1a shows the location of Mersea Island on the South Eastern coast of

England while the map in 1.1b shows the position of the island in relation to the coast of

Essex. The following maps, (1.1c) and (1.1d) zoom in on this geographical area so that

the Island and the surrounding locations can be seen in more detail.

14

1.1 c) From http://www.multimap.com/maps/#t=l&map=51.84169,0.86927|12|4

Mersea Island

With respect to the Island itself, the population of West Mersea is much denser than the

population of East Mersea and this is highlighted in 1.1d by the denser road grid on the

western side.

1.1 d) From http://en.wikipedia.org/wiki/Image:Mersea_islandmap.jpg

15

1.2

Access

The Island is not permanently accessible from the mainland. Travel to and from the

Island is hindered by the lunar tides which routinely flood the only access road, the

Strood (from the Saxon strod meaning ‘marshy land’). Morant notes in his book A Short

History of Essex (cited in Karbacz (1999:41)) that, in his time (the late 1700s), the Strood

was only passable eight hours a day. Even though current native Islanders generally feel

that the modern tides are higher than they used to be, such dramatic cut-off times as

experienced in Morant’s day no longer occur. However, regular travel to the mainland

(or indeed simply from West Mersea to East Mersea) was not widely available or usual

until twentieth century transport technology was introduced. Thus, many older

generation Islanders cannot remember leaving the Island until their late teens or older.

Indeed, off-island travel was viewed as an experience and even contributed, in one case,

to the allocation of the type of personal nickname so prevalent in this type of small

community. In this case, the nickname ‘Foreigner’ was given to a man who sailed across

to Whitstable in Kent to work on their oyster beds for only a short period and was judged

to have come back to the Island talking with a different accent.

1.3

The Island’s Geographic Development

On many Island maps, Mersea is labelled as consisting of two parts: East Mersea and

West Mersea. This is also reflected in the road system since, when you first enter the

Island from the Strood, the road branches in each direction. The eastern side remains

largely agricultural and has seen little increase in its population over the last 100 years

(see figure 1.2 below). Conversely, the western side has witnessed a dramatic increase in

16

its permanent population (660 in 1801 to 6926 in 2001) due to numerous housing

developments, and continues to experience a large transient population from increased

levels of tourism, especially during the summer months. These tourists include those who

only come to the Island for the day, often to visit the popular seafood restaurant, The

Company Shed. However, Mersea also has five holiday and caravan parks for those who

wish to stay longer, as well as a number of independent bed and breakfast and holiday

properties.

Although the division between East and West Mersea is not significant in modern Island

life, due to advances in motorised transport and loss of the East’s primary school around

the middle of the last century, this has not always been the case. Sourcing the Anglo

Saxon Chronicle, Karbacz (1999) explains that retreating Danish invaders around the

ninth century AD settled on Mersea before advancing once again up the Rivers Thames

and Lea. During this time, a moated area of approximately six acres is thought to have

been constructed around the East Mersea area, effectively enclosing their encampment.

Less than two centuries later, when the Saxon dynasty was restored prior to the Norman

Conquest, further reinforcement of the East-West divide took place when the Manor of

West Mersea (which included parts of the parish of Peldon and parts of Fingringhoe on

the mainland, highlighted in Figure 1.1c) was inherited by the Crown, while the Manor of

East Mersea was held privately (Karbacz (1999:16)). This separation is still visually as

well as psychologically maintained between the parishes through the Deremy Stone

which marks the boundary point between the two.

17

Another notable effect on the Island’s demographic structure came during the sixteenth

and seventeenth centuries when many Dutch and French refugees fled Continental Europe

and settled across East Anglia. The impact of these migrants is still reflected through a

great number of, albeit anglicised, local surnames and it is widely believed that the many

Island families bearing the name ‘French’ are descended from those whose names were

beyond the locals’ linguistic comprehension and capabilities.

More recently however, with the advent of World War ІІ, Mersea became home to not

only a range of military personnel (as with World War І) but was also opened up to many

war-time evacuees for varying lengths of time. The origin of the evacuees was quite

diverse, although local recollections claim the majority were from London. The extent of

contact with the evacuees varied from person to person but their leaving, at the end of

World War II, also coincided with the start of the steady population rise on the Island as

land was sold and new properties were built that became widely available to native

Islanders and Mainlanders alike.

18

Popluation of Mersea Island based on census data from 1801 to 2001

0

1000

2000

3000

4000

5000

6000

7000

18

01

18

11

18

21

18

31

18

41

18

51

18

61

18

71

18

81

18

91

19

01

19

11

19

21

19

31

19

51

19

61

19

71

19

81

19

91

20

01

Date

West Mersea

East Mersea

1.4

Modern Mersea

The following graph shows how the population of West Mersea has dramatically

increased, especially since 19616 with the population in 2001 recorded at 6926. This is in

stark contrast to the East Mersea population which, due primarily to building restrictions,

never rises above 331, which was its peak in 18417.

1.2

The population of Mersea Island, based on Census data from 1801 to 2001

However, it is also important to look at the age group representations of the Island

population. The following data show the population breakdown into discrete age groups

for West Mersea, Colchester (Mersea’s closest large town) and the East of England as a

whole:

6 Note: the apparent sharp rise pre-1961 is due to there being no records for 1941.

7 Note: there is no East Mersea data for 2001 due to its reclassification to a separate Census ward.

19

1.3 The 2001 population breakdown of West Mersea, Colchester and the East of England.

a) The population numbers

West Mersea Colchester East of England

Aged 0-9 669 19288 665408

Aged 10-19 695 19599 668680

Aged 20-29 510 22072 637383

Aged 30-39 805 24143 827040

Aged 40-49 844 20179 728588

Aged 50-59 1113 20420 707870

Aged 60-69 909 13319 510554

Aged 70-79 864 10434 407903

Aged 80-89 449 5389 198925

Aged 90+ 68 953 35789

Total 6926 155796 5388140

b) The 2001 population percentages


Aged 0-9 9.66 12.38 12.35

Aged 10-19 10.03 12.58 12.41

Aged 20-29 7.36 14.17 11.83

Aged 30-39 11.62 15.50 15.35

Aged 40-49 12.19 12.95 13.52

Aged 50-59 16.07 13.11 13.14

Aged 60-69 13.12 8.55 9.48

Aged 70-79 12.47 6.70 7.57

Aged 80-89 6.48 3.46 3.69

Aged 90+ 0.98 0.61 0.66

20

c)

The 2001 Population of West Mersea, Colchester and the East of England according to

the percentage of people within each age group

0

10

20

30

Aged 0-9 Aged 10-

19

Aged 20-

29

Aged 30-

39

Aged 40-

49

Aged 50-

59

Aged 60-

69

Aged 70-

79

Aged 80-

89

Aged

90+

Age of Population

Pe

rce

nta

ge


One aspect which is immediately noticeable is that Mersea has a relatively low population

of those under the age of 49 and a higher percentage population for all age groups of 50

years old and above when compared to nearby Colchester and the East of England as a

whole. One reason for the lower levels of young adults, especially in the 20-29 age

group, could be a result of increases in the number of people able to attend university or

training colleges who, as a result, decide to move away from home to their university

town or city. It may also reflect the lack of local unemployment and unavailability of

many Island homes to first time buyers in this age group. Indeed, it was a common

complaint amongst many of my younger informants that house prices on the Island were

too high for them to consider moving out of home to another Island property. The loss of

Islanders from the younger age brackets can also be seen in the migration figures from

previous years.

21

1.4 The Net Migration Statistics, from the 2001 Census, for the 15-24 year old age group

in the West Mersea

a)

Population Turnover Rates

Net Change; Aged 15-24

Persons per 1000

West Mersea

July 2001 to June 2002 33

July 2002 to June 2003 -36

July 2003 to June 2004 -23

July 2004 to June 2005 -30

July 2005 to June 2006 -14

July 2006 to June 2007 -8


July 2008 to June 2009 -48

b)

Graph to show the net population change in Islander numbers aged 15-24 between July

2001 and June 2009

-60

-50

-40

-30

-20

-10

0

10

20

30

40

July 2001 to

June 2002

July 2002 to

June 2003

July 2003 to

June 2004

July 2004 to

June 2005

July 2005 to

June 2006

July 2006 to

June 2007

July 2007 to

June 2008

July 2008 to

June 2009

The Year of Data Representation (based on the 2001 Census)

Pers

on

s p

er

1000

It can be seen how the net figures for the 15-24 year old age group (calculated by

subtracting outflow numbers from inflow numbers) show quite significant negative

values, which peak in the 2008-2009 period, giving a difference of -48 persons per 1,000.

22

This indicates that the loss of Islanders before the age of 25 is a regular occurrence,

contributing to the lower population figures for these age groups. This is in direct

contrast to the migration figures for those in the next age group up - the 25-44 year olds.

The net figures for this group demonstrate consistently positive figures between 2005 and

2009, which suggests that, even though the Island is losing its population below the age

of 25, it is gaining through in-migration with those aged over 25. However, it is uncertain

how many of these residents are employed on the Island or indeed how many children of

original mainland families attend West Mersea school.

1.5 The Net Migration Statistics, from the 2001 Census, for the 25-44 year old age group

a)

Population Turnover Rates

Net Change; Aged 25-44

Persons per 1000

West Mersea









23

b)

Graph to show the net population change for Islanders aged 25-44 between July 2001 and

June 2009

The loss of Islanders from younger age groups may be facilitated, at least in part, by the

greater number of opportunities available to modern Islanders to expand their social

networks beyond the Island from an early age. In contrast to older generations (who were

educated to the ages of 14 or 15 in the local school on the Island), the introduction of

compulsory schooling to the age of 16 has lead to Islanders leaving Mersea for their

secondary education, as Mersea school only provides primary education. In addition, if

tertiary education is sought, this too must be completed off the Island. This automatically

introduces Island children to mainland children and so social network ties can become

established across both areas. This in turn provides greater opportunities and motivation

for personal geographical mobility.

Another important factor in the socio-economic development of the Island is that of local

industry. For those belonging to previous generations, it was unusual for someone to be

-10

-5

0

5

10

15

20

25

30

35

July 2001 to

June 2002

July 2002 to

June 2003

July 2003 to

June 2004

July 2004 to

June 2005

July 2005 to

June 2006

July 2006 to

June 2007

July 2007 to

June 2008

July 2008 to

June 2009

The Year of Data Representation (based on the 2001 Census)

Pers

on

s p

er

1000

24

employed off the Island, especially with respect to their first jobs. Therefore, local

Mersea-based jobs were considered the norm and not the exception. However, Jay,

writing in 1996, notes that “at the present time, only one local shop remains trading under

its family name” (1996:1).

That being said, the following data show that Mersea is still a place for established

businesses, even though the original families may no longer be involved in their running.

1.6 Data Demonstrating the Age of Businesses registered in 2010 with a comparison

between West Mersea, Colchester and the East of England

a) The total number of businesses


Less than 2 Years Old 35 875 30890

2 to 3 Years Old 40 840 30480

4 to 9 Years Old 90 1740 59640

10 or More Years Old 155 2535 92630

Total - all VAT and/or PAYE Based Enterprises 320 5990 213640

b) The percentage of businesses by age and statistical region


Less than 2 Years Old 10.94 14.61 14.46

2 to 3 Years Old 12.50 14.02 14.27

4 to 9 Years Old 28.13 29.05 27.92

10 or More Years Old 48.44 42.32 43.36

25

c)

The graph showing the percentage of businesses by age and statistical region, based on

2001 Census data

0

10

20

30

40

50

Less than 2 Years Old 2 to 3 Years Old 4 to 9 Years Old 10 or More Years Old

Age of Business

Perc

en

tag

e


From this 2010 data, we can see that, with respect to ‘new’ business, Mersea is below the

percentage for both Colchester and the East of England. Conversely, it is very much

leading in the category of businesses over 10 years old. This suggests that, even though

traditional families have ceased to be involved in the long-term business which they

originally established, there is still relative stability in local trade.

It is also notable that the type of employment has become a lot more diverse within the

modern Mersea population. Historically, Mersea’s main trades were associated with the

sea and the land, and therefore it was common for Islanders to go into trades associated

with fishing (and particularly oyster farming) and agriculture, as well as family trades

such as building and plumbing.

However, according to Census data from 2001, only 0.64% of Islanders were involved in

the fishing industry, compared with 0.02% across the East of England. In addition, only

26

1.61% of the working population was registered as being involved with agriculture,

hunting and forestry. This results in two of the historically important employment types

being in the bottom four employment areas overall, as the table below demonstrates:

1.7 Data representing types of employment for residents of West Mersea, Colchester and

the East of England

a)


Agriculture; hunting; forestry 48 1021 49009

Fishing 19 33 637

Mining & quarrying 7 98 5457

Manufacturing 368 8450 373155

Electricity; gas and water supply 11 392 16223

Construction 304 5843 196461

Wholesale & retail trade; repair of motor vehicles 551 13013 445887

Hotels and catering 112 3160 107418

Transport storage and communication 189 5009 191252

Financial intermediation 175 5159 149883

Real estate; renting and business activities 309 8826 343264

Public administration and defence 120 5750 133066

Education 249 6301 189274

Health and social work 335 8799 249776

Other 183 3272 128616

Total - All people aged 16-74 in employment 2980 75126 2579378

27

b) Percentages of types of employment, ranked according to the percentages of Mersea

residents


Wholesale & retail trade; repair of motor vehicles 18.49 17.32 17.29

Manufacturing 12.35 11.25 14.47

Health and social work 11.24 11.71 9.68

Real estate; renting and business activities 10.37 11.75 13.31

Construction 10.20 7.78 7.62

Education 8.36 8.39 7.34

Transport storage and communication 6.34 6.67 7.41

Other 6.14 4.36 4.99

Financial intermediation 5.87 6.87 5.81

Public administration and defence 4.03 7.65 5.16

Hotels and catering 3.76 4.21 4.16

Agriculture; hunting; forestry 1.61 1.36 1.90

Fishing 0.64 0.04 0.02

Electricity; gas and water supply 0.37 0.52 0.63

Mining & quarrying 0.23 0.13 0.21

Even though the percentages for the number of people employed in the fishing industry

on Mersea is greater than that of Colchester and the East of England, this does not

necessarily indicate continued strength in this trade. Colchester is inland and so it would

be unusual for people to be employed in such an activity and the data for the East of

England region represents an aggregate of all areas, both inland and coastal, the former

having greater populations through which the data is influenced. Therefore, it is

important when using data such as this that we look at it on a community level and

compare similar communities, as opposed to automatically comparing communities in the

same geographical regions. For example, the Norfolk district of Great Yarmouth and the

Suffolk district of Waveney (which includes Lowestoft) are both historically noted as

areas associated with the fishing industry. Data from the 2001 census show that only

0.3% of the 46,001 Waveney population work in fishing while this percentage is reduced

to 0.03% of the 37,584 of the Great Yarmouth working population. Therefore, even

though the Mersea employment percentages for this occupation are lower than we might

expect, they are higher than both Great Yarmouth and Waveney districts.

28

However, even though we may want to claim that there is a decline in traditional local

industry, additional data show that local travel is more common than long distance

commuting:

1.8

Percentage of the West Mersea workforce population (aged 16-74) according to distance

travelled to work, from the 2001 census

0

10

20

30

40

50

Works

mainly at or

from home

Less than

2km

2km to less

than 5km

5km to less

than 10km

10km to

less than

20km

20km to

less than

30km

30km to

less than

40km

40km to

less than

60km

60km and

over

No fixed

place of

work

Working

outside the

UK

Working at

offshore

installation


This data set demonstrates the distance travelled to work by Islanders aged 16-74 who are

in some type of employment. We can see that 12% of Islanders work from home and

18% travel less than 2km (1.2 miles) to work – this latter group will be working on the

Island, travelling around West Mersea. However, the most common distance travelled to

work is 10km – 20km (6.2 – 12.4 miles), representing just under 40% of the Island’s

workforce (39.56%). This is the distance which would represent travel to Colchester and

its surrounding areas. Therefore, it seems that Islanders, who are integrated with

mainland varieties from an early age through school, remain tied to mainland in their

professional activities.

29

1.5

Summary

The data presented in this chapter suggest a number of influences on the Mersea

population. The population figures demonstrate a continuous rise in West Mersea’s

residential population over the last century, and that over half of this population is now

professionally employed off the Island. The migration figures from 2001 to 2009 also

suggest that, even though the population is increasing, the main age group which is

experiencing diminishing numbers is the 15-24 group which helps promote the elasticity

of family ties, as members become more scattered, and lessens the Island’s adolescent

population. Conversely, the migration statistics for the 25-44 age group suggest that the

residential population is being bolstered by in-migrants from this age group. The contact

between Islanders and Mainlanders is also promoted through short term visitors, tourists

and holiday makers. This transient population, which is at its peak in the summer

months, adds to the level of contact between dialect varieties. Thus, dialect contact is

being promoted in two ways:

1. Connections with the mainland have become greater and more extensive. As a

result, more Islanders are travelling off the Island for education, work, socialising

etc.

2. More and more people from the outside are coming to the Island. This is being

achieved through tourism and house-moving.

30

Chapter 2

Methodology

31

2. METHODOLOGY

This chapter will discuss the methodology used for this study. As previously outlined,

this study is concerned with the analysis of the dialect spoken on Mersea Island. There

has been no systematic sociolinguistic research carried out on the Island, and it has only

been represented by one speaker from East Mersea in the Survey of English Dialects in

past dialect surveys. Therefore, in order to study the characteristics of the Mersea dialect,

spoken data from Islanders needed to be obtained. In addition, in order to investigate

whether phonological change has taken place, and if so what direction has this change

taken, data needed to be collected from a range of sources so that both an apparent time

and a real time comparison could be made. However, since the focus of this study is the

nature of the modern Island dialect, and how contact with dialects from off the Island may

have affected the dialect of those on the Island, only native Islanders were considered

appropriate for this study.

Therefore, included in this Chapter will be discussions relating to the selection of

informants and the interview structure. In addition, processes involved in the analyis of

the data with respect to token extraction, coding and phonological distribution of the

variables will also be presented.

Finally, this Chapter will provide a brief overview of the primary historical sources which

were used to inform the historical development of each variable in the geographical

region under study.

32

2.1

Data Collection

The following sections will outline the steps taken while collecting data of the Mersea

Island dialect. This was done through a number of informal, casual conversation

interviews (aimed at collecting as relaxed and unmonitored speech as possible) from a

range of Mersea Island natives. These informants were chosen to represent two distinct

age groups so that different stages of the Island’s social and historical development were

represented. In doing this, it is not only the dialectal stages which may be reflected, but

also any changes in attitudes towards the Island community and historical perspectives or

memories (particularly from the oldest speakers) which may not have been recorded

elsewhere. Access to the very oldest members of the community, who would have

represented the dialect in the early twentieth century, proved nearly impossible.

However, with the discovery of an audio archive held by the local museum, a small group

of speakers representing this period of Island life became available for analysis. The

nature of these recordings will also be discussed in greater detail below.

2.1.1

The Interviews

Approaches to Sociolinguistic data collection have taken a number of forms over the

years. These varying methods have included written questionnaires, rapid anonymous

surveys, the reading of word lists and pre-constructed written passages, as well as

structured and unstructured interviews (see, for example, Milroy and Gordon (2003) for a

review of these methods). The method I adopted for the study of Mersea Island English

(MIE) was that of casual, unstructured interviews. Thus, the goal of each interview I

conducted was to capture the type of spontaneous fluent speech associated with the

33

variety’s vernacular. These interviews were always conducted in the informants’

households in a room of their choice. Milroy and Gordon (2003:57-58) note that a

significant advantage of this approach is that, due to the lack of a predetermined structure,

such as that associated with survey questionnaires, the successful interviewer is able to

elicit more extended stretches of conversational (and thus unscripted) speech from the

informant.

However, as is often the case with interaction of this nature, the sociolinguist’s goal of

achieving true unmonitored speech on behalf of the informant is subject to the effects of

the observer’s paradox. This is outlined by Labov (1972) with respect to the aims of

linguistic research in the community as a means of discovering “how people talk when

they are not being observed; yet we can only obtain these data by systematic observation”

(1972:209). In addition to diverting an informant’s attention from the interview situation

through distractions such as telephone conversations, Labov (1972) indicates that the

effects of the paradox may be overcome by involving the subject in questions and topics

which provoke strong emotions. Even though Labov specifically introduces the notion of

the “danger of death” scenario as a prompt for such emotive responses, this approach has

not been proved effective or appropriate for other communities under observation.

Milroy and Gordon (2003:65-66) note that enquiries of this sort in the context of Belfast

interviews elicited quite matter-of-fact accounts of dangerous experiences rather than the

emotional responses achieved by Labov. Thus, the lack of a reliable cross-community

strategy led Milroy (1987) to write that “direct interviewing, however informal on the part

of the field worker, is an uncertain means of gaining access to the vernacular [and] that

the interaction between language and situation is too complex and too little understood

for an interviewer to be able to manipulate it reliably” (1987:26).

34

Regarding the interviews used for this study, I attempted to elicit relaxed conversation

and thus minimise the effects of the observer’s paradox by exploiting both my status as a

native Islander and existing community ties, as well as utilising information gained by

observing the surrounding environment and information gained from previous interviews

with acquaintances. The former allowed me to initiate relevant conversation about local

issues or acquaintances we may have had in common (specifically the person or persons

through whom I was recommended to them), while setting up the necessary recording

equipment in order to help divert attention away from my actions and the presence of the

equipment. In many cases, this was not immediately necessary as I was able to set up

equipment in relatively inconspicuous places while the informants were making cups of

tea. It must be noted at this point that, prior to any equipment activation (usually when I

was first admitted to the home), the informants were explicitly asked if they would permit

me to record them and thus, I provided them with an opportunity to refuse. I also

provided information about my studies and welcomed any questions they had regarding

my work. This procedure itself often generated conversational openings such as

discussions on technological advances. In addition, each informant was asked at the

conclusion of the interview to sign and date the cover of their mini-disk case as evidence

of their consent and also if they would like to be informed of any results from my work.

The use of visual information gained through the location of the interview also provided a

great source of relevant topical conversation. By observing and commenting on, for

example, the type of books in a bookshelf or pictures and paintings on the wall, I was able

to unobtrusively enquire about an informant’s hobbies and family. By introducing and

manipulating conversational topics in this way, I was able to learn which topics appeared

to relax an informant or generate enthusiasm on their part and I could then exploit these

moments.

35

The data utilised for this research focussed on the speech of twenty-eight informants.

The data for five of the male informants came from cassette tapes held by the local

museum and formed part of their archive (see below for more details). These interviews

ranged from 20 to 45 minutes and were recorded in 1978/9. The interviews conducted by

myself (which account for the data pertaining to both the Older and Younger age groups)

ranged from approximately 60 to 165 minutes and were conducted between January 2006

and March 2007. The recording equipment used was a Sony Hi-MD Walkman MZ-RH10

and its appropriate mini-disks. These recordings were then transferred from mini-disk to

ordinary CD disks for audio analysis.

2.1.2

The Informants – A General Overview

Designated Age Group Male Female

Museum 5 0

Older 7 7

Younger 5 4

Total - 28

Twenty eight speakers were chosen, based on the criterion that they had lived on Mersea

Island for their entire lives. Thus, new in-migrants (that is those residents who had

moved to the Island and were not life-long residents) were excluded from this study. Due

to the limited nature of data from East Mersea informants in my corpus, and to achieve

consistency within and across each data set, all speakers chosen for this study had spent

their lives living on the western side of the Island.

36

Data from the Museum archive was in the form of audio cassette tapes. These were

originally produced through the local Lions charity as a means of creating an audio

magazine for Islanders. They included news updates and reports on local events as well

as interviews with Islanders who spent time either talking about a particular topic or about

their life in general as they grew up on the island.

Informants for my own interviews were recruited by one of three methods:

1) THE SNOWBALL METHOD:

This method, as described by Milroy and Gordon (2003:32), takes advantage of

the social networks of participants. A particular advantage of this method, where the

researcher simply asks informants to recommend others whom they feel would be willing

to participate in the study, is a reduction in the likelihood of potential subjects declining

to take part. In doing this, the researcher can begin to shed the status of complete outsider

for one of a friend-of-a-friend, which has been effectively used by, for example, Milroy

(1987) in the study of Belfast communities.

2) PERSONAL INTRODUCTION BY A THIRD PARTY:

Many useful contacts were made throughout the older generation within the

Mersea community when I was invited to help as a volunteer for a local club which

organises activities for those who are partially-sighted. I was introduced to the club

members who had been told about my work and I spent some time talking to them and

giving them my contact details. Many who were not life-long Islanders were still happy

to pass my details on to people they knew.

3) PERSONAL RELATIONSHIP WITH THE RESEARCHER:

These informants make up the smallest group across the Mersea corpus with only

3 informants (all among the younger generation) having a prior relationship with myself.

37

2.1.2.1

The Museum Speakers

The data-set which has been given the ‘Museum’ classification is named as such since the

recordings were sourced from archives held at Mersea Island’s local museum.

Since 1978/9, Mersea Island Lions Club has produced a talking magazine which is

distributed on ordinary cassette tapes and aims to provide a combination of local news

and event updates, as well as interviews with local people who discuss particular topics

(such as the development of local business, transport or simply reminiscences of times

past). The interviews, to which I had access, were conducted mainly at the local council

offices (though occasionally, due to the age and presumably limited mobility of the

participants, they were conducted in the home). Biographical information of those being

interviewed was mainly unavailable, and, although an online description of individual

tapes is in the process of being constructed, many of the entries only have the order of

contents and the name of the interviewee.

Once I was provided with digital copies of the oldest recordings, I methodically went

through them in order to judge which speakers were appropriate. During this time, I also

made notes on recordings that I did not use, so that I too could work with the museum to

update the online site. Therefore, from this period, I chose interviews of five male

speakers (there were no female speakers interviewed during this period who were native

Islanders) and these speakers were born in the years 1883, 1890, 1910, 1916 and the mid-

1920s (- the last of which did not give a specific birth year within the interview but a

rough guess could be made, based on the information provided in the dialogue). It could

be argued that the dates of birth are too widely spread to be combined within the same

38

speaker group, as there is a difference of around 40 years. However, with limited data

available for each speaker, individual speaker analysis would not have elicited enough

tokens for a reliable individual comparison.

Each of the five informants was a native Islander. However, due to the time period in

which they grew up, each member of this group served at some point in the armed forces

and thus travelled away from Mersea. However, each returned after their period of

service had ended and returned to Island employment.

2.1.2.2

The Older Generation

Due to the nature of the educational system on Mersea until the mid-1900s, all older

speakers were educated on the Island until the age of either 14 or 15, depending on their

era, and none received any additional education after this point (with the exception of one

female who went on to train as a nurse). After this time, each older male in the sample

was employed on the Island in local trade. Conversely, two older women were initially

employed off the Island once they had left school but returned to Mersea employment in

later life.

2.1.2.3

The Younger Generation

In order to retain some consistency regarding educational location, I attempted to

interview those informants of the younger generation who had been schooled at West

Mersea primary school. This was achieved for 8 out of 9 informants. The exception to

this was one female who attended a local Colchester private school for part of her primary

39

education. In addition, all informants from this data group attended a local secondary

school in nearby Monkwick and either the local Sixth Form College or local Institute

(both located in central Colchester) for their tertiary education. However, one other

female spent years 7 and 8 (thus aged 11-13 years old) in a Sussex boarding school before

returning to attend the Monkwick school for the remainder of her secondary education.

2.2

Social Factors

This section includes a brief overview of some theoretical issues involved in the

classification and categorisation of informants.

2.2.1

Class

In order to achieve valid comparisons within and across sociolinguistic studies,

sociolinguistic factors must remain as consistent as possible. However, maintaining such

uniformity across research is not only problematic but unrealistic. By way of illustration,

the calculation and constructions of abstract socio-economic class categories through

methods such as the social class index utilised by Trudgill (1974) in Norwich, which was

designed “to measure objectively the social class and status characteristic of the Norwich

sample” (1974:35), may only be applicable to a particular subsection of one society.

However, this cannot accommodate factors such as persons of the same class having

different statuses if they reside in a location which assigns a different class structure

(Milroy and Gordon 2003:44). For example, a builder may hold a different status and be

assigned to a different class in two different communities depending on how each

40

community values that profession, or continued family lineage in a particular geographic

area may be held in higher or lower esteem depending on the society.

The calculation of class on Mersea Island indeed faces such difficulties, as evaluations

appear subjectively dependent upon the subculture to which a person belongs. Thus, it is

from my own experience and impressions from those Islanders with whom I have had

contact, that Islanders whose families are historically established as native families are

held in higher esteem by those from similar local lineage (as opposed to those who have

recently moved to the Island). Therefore, it seems that family history is a far more

significant marker of higher social status than occupation for these residents. This is

perhaps in part due to the perception of professions such as building, shipwrighting and

fishing being derived from ‘Old Mersea’ when it was common for sons to adopt the

father’s, or family’s profession. This type of evaluation (although not as strongly put

forward) was still evident among many of the younger generation of informants, as each

showed an awareness and ability to name prominent Island family-names. Therefore, due

to the subjectivity of these views and the inherent difficulties with calculations of social

class, this was not considered as a social factor for this study. Furthermore, no speaker

recorded during the process of data collection was excluded on the basis of their apparent

class or social standing. For example, no member of the Older speaker group attended

university and all represented similar professions (for example, shipwright, oyster

fisherman and builder). In addition, all members of this group were involved in some

kind of Island activity, from being a organiser at the Church to a member of one of the

local committees.

41

2.2.2

Age and Gender

Age and gender as social variables lend themselves, on the surface, to straightforward

classifications through the categorisation of numerical age and the binary distinction

between males and females, respectively. However, age and gender as social constructs

have been treated more intricately than simple assignment of arbitrary categories. By

way of illustration, Eckert (1998), presenting data from her Belton High sample, posits

that gender be treated as a continuum, since the concept of gender appears to be

intertwined with additional aspects of social identity. Similarly, Eckert (1997) asserts that

age be treated in the form of life stages, which primarily distinguishes childhood from

adolescence from adulthood and explains that age systems “serve to mark not only an

individual’s progress in the life trajectory, but the individual’s progress in relation to

societal norms” (1997:155). Regarding the Mersea sample, informants have been

categorised simply as male or female, since a more detailed in-depth analysis of this

social construct is beyond the scope of this thesis. Similarly, speakers have simply been

assigned to the classification of Older or Younger generation subsets. The Younger

generation of speakers range in age from 18 to 26 years, while the Older generation range

from 58 to 75 years of age.

42

2.3

The Variables – Extraction, Analysis and Classification

All interviews were analysed from start to finish resulting in approximately 32 hours of

data. The analysis of the variables (au), (a) and () were conducted as an

impressionistic or auditory analysis rather than an instrumental analysis8. This was due,

in part, to a lack of consistent recording quality across the interviews, especially the

interviews of the Museum speakers. Also, the focus of this thesis is the representation of

sociolinguistic variation within a phonological model. Therefore, detailed phonetic

descriptions would not be as valuable to the model presented in Chapter 10 as the

perceived phonological categories of the surface variants. Once tokens were extracted

and transcribed, random samples were selected and reanalysed at a later date to ensure

consistency with the original transcription. On the odd occasion where the two

transcribed forms differed, the token was re-analysed for a third time and the majority

form was adopted.

The following sections outline the procedure through which the coding and classification

of data was carried out for the purpose of the analysis.

2.3.1

Linguistic Constraints

Linguistic constraints, or linguistic factors, represent any internal factor which may

correlate or influence sociolinguistic variation. Therefore, with respect to this study,

certain phonological constraints may restrict the distribution of each variable (see below),

while aspects such as specific phonological environments may correlate with the

8 Even though these three variables are the focus of this study, the full (stressed) vowel inventories for two

Mersea Island speakers can be found in Appendix E

43

production of some variables more than others. Thus, all variables were coded according

to the same broad phonological categories:

1. Syllable Type

2. Preceding Phonological Environment

3. Following Phonological Environment

while the following linguistic aspects were taken into account:

4. Prosodic Factors – stress

5. Specific Lexical Items – the case of LIKE, ISLAND and BOY (see the relevant

chapters for details)

2.3.1.1

Syllable Type

The classification of syllable type was a binary choice between those which were open

and those which were closed. Therefore, distinctions such as the following could be

made:

Open – Closed

(au)

allow – loud

houses – house

(a) fly – flight

fighter – slightly

() boy – boys

toilet – jointed

44

2.3.1.2

Phonological Environment

With respect to phonological environment, tokens were coded according to actual, surface

pronunciations as opposed to underlying representations. Therefore, by way of

illustration, the underlying /t/ in lighter could have surfaced (and thus be classified) in

one of three ways – [t], [] or [], depending on the speaker’s pronunciation. Word-initial

or word-final tokens were classified as having no preceding or following environments on

the word level, respectively, and, in the segmental analysis, are represented by the

category ‘None’.

However, with respect to following phonological environment, it must be noted that all

three variables have distributional restrictions, with not all potential following

environments possible in the dialect. These restrictions (following Hammond (1999) who

considers the distribution of American English phonemes in mono-morphemic words),

together with any additional considerations for each variable, will be discussed in more

detail below. However, it is worth noting that Mersea Island English is a non-rhotic

dialect9, so the possibility of following /r/ has been omitted from the discussions.

2.3.1.2.1

The Distribution of /a/

The diphthong /a/ has very restrictive co-occurrence rules with respect to following

segments in English. Indeed, Hammond (1999) illustrates how /a/ does not occur before

any word-final non-coronal consonant. Thus, /pat/ is possible but not */pak/, and it is

9 This is referring to the dialect in the present day. There is evidence of r-colouring in the Survey of

English Dialects and in the speech of an East Mersea informant (whose data is not included here since this

study focuses on those from West Mersea) who was born in 1917.

45

only final /st/, /nt/ and /nd/ cluster which are able to follow it word-finally (for example,

joust, fount and bound).

With respect to word-medial contexts, Hammond suggests that /a/ can occur before /b/

in baobab, before /f/ in dauphin and before // in caoutchouc (1999:121, 123 and 126,

respectively), although these pronunciations seem slightly idiolectal and prone to inter-

speaker variation. Thus, word-medial /a/ has greater restrictions, since it is only found

preceding /t d z/ and the clusters /st/ and /nt/.

2.3.1.2.2

The Distribution of /a/

The distribution of the diphthong /a/, while having a wider distribution than /a/, remains

restricted with respect to which phonological segment can immediately follow it.

The distribution of /a/, in mono-morphemic words, in contrast to /a/, does permit /a/ to

occur before non-coronals, such as labial /m/ and velar /k/, for example climb and hike.

However, /a/ does not co-occur with a following / / in word-final position. With

respect to clusters, in a similar fashion to /a/, the only clusters which are phonologically

permitted to follow /a/ in English are /st/, /nt/ and /nd/ (for example, heist, pint and find).

However, /a/ may also precede /ld/ clusters in words such as wild.

Regarding medial segments, a similar pattern emerges, with the exception of /a/ medial

sequences (such as in tiger). A following dorsal // and the palatals // and // are still

disallowed, while the only medial cluster permitted to follow /a/ is /st/ (such as the loan

word maestro).

46

2.3.1.2.3

The Distribution of //

The distribution of // in monomorphemic words is similar to that already described for

/a/. Following Hammond (1999), we can see that, word-finally, // can also not appear

before non-coronals (though Hammond (1999:111) does suggest coif as a possible /f/

sequence) and the only final clusters which follow // are /st/ and /nt/. However, this

does not take into account the colloquial British English word ‘oik’10

which sees //

preceding a velar consonant.

Word-medially, however, // may precede all oral stops (except /g/), only the coronal

fricatives /s z/ and the approximant /l/. It is not permitted to precede nasals or affricates

in medial position. In addition, it is only following medial /st/ clusters (in words such as

cloister) which are allowed.

2.3.1.3

Stress

Stress placement was not considered as a linguistic variable as all the tokens appear in

stressed syllables. With lexical words however, with respect to /a/, this variable may

also appear in the function words I, my, myself and by in which reduction to [ ~ ] is

common due to their unstressed position in an utterance. For this reason, they have not

been included in the current analysis which instead is focussed primarily on a comparison

of /a/ variants in lexical words as opposed to function words.

10

which is described by the New Penguin English Dictionary as ‘a vulgar, loud, or otherwise obnoxious

man; a yob.” (2000). The New Penguin English Dictionary, Penguin Books.

47

2.3.1.4

LIKE

A further consideration when analysing /a/ is its presence in the token ‘like’. D’Arcy

(2007) identified nine different classifications of like under two broader headings relating

to whether its function is Grammatical or Discursive in nature (2007:392, see Chapter 5.6

for further details).

Drager (2009) used these classifications as a basis for her analysis of like in the speech of

female high school students in New Zealand. She found that the function of like had a

strong correlation with the type of variation produced and, so, the analysis presented here

coded the various functions of like separately and will also examine the different

functions of like apart from the main /a/ analysis.

2.4

The Historical Data

This section will briefly look at the historical data sources which were used to gain a

historical insight to the past developments of each of the variables.

48

2.4.1

Ellis (1889)

Throughout the late nineteenth century, Ellis compiled an extensive corpus of ‘dialect

renderings’ from across the Anglophone parts of England, Ireland, Scotland and Wales

using a three-part methodology:

I. Comparative Specimen (cs) – this involved a reading passage, the purpose of

which was to obtain “dia[lect] renderings of familiar words in various connections and

some characteristic constructions” (1889:7*).

II. Dialect Test (dt) – an additional short passage designed to elicit almost all of the

word categories in the classified word list (see below).

III. Classified Word List (cwl) – an extensive collection of items, which were

subdivided under the headings Wessex and Norse, English and Romance according to

“the vowel of the original language in what corresponds to the accented syllable in

received English” (1889:16*).

Ellis allocated Essex to his Eastern Division and describes this division as follows:

“A closer resemblance to received speech than in any other div[ision]. It is the

region from which rec[eived] sp[eech] was taken, and contains the greater part

of London. The pron[unciation] is, however, not quite uniform, but the

differences are so slight that it has been found extremely difficult to obtain

satisfactory information” (Ellis 1889:188).

Within this division, Essex is grouped with Hertfordshire, Bedfordshire, Huntingdonshire

(in present day Cambridgeshire) and mid-Northamptonshire to form a Mid Eastern dialect

district, whose geographical character is described as long and straggling. He also notes

that, even though the differences between the north of Northamptonshire and the south of

Essex are marked with little or no connections, “I have found it impossible to divide the

district….and have felt it best to consider the different counties involved as forming

‘varieties’” (Ellis 1889:196).

49

As well as a general overview of each division, Ellis gives a description of each county

area within that division, highlighting the prominent dialect characteristics across each of

the methodological approaches. Therefore, it may not be the case that each county

description features the same sociolinguistic variables. Due to this, the assessment of

(), () and () will rely on any descriptions presented as part of the overview of the

Essex region. However, since Ellis used a particular palaeotype from transcribing and

describing sounds and variants, the translation of these symbols to their equivalent

modern IPA symbols by Eustace (1969) will be employed in this thesis.

2.4.2

Wright (1905, reprinted in 1968)

The construction of Wright’s dialect grammar aimed to highlight and describe the main

characteristic features of all British dialects. To do this, it was originally intended to be

based upon non-literary words, as he proposed that their phonological structure would

have been less influenced by literary language (Wright 1968:III). However, due to the

restricted geographical spread of genuine dialect words, Wright’s index of 2,431 items

was mainly based upon words which were represented by both the literary language and

the spoken dialects themselves. The classification of dialects by Wright follows those of

Ellis (discussed above) as he subdivided the Eastern Division into five groups, of which

the Mid Eastern group contained most of Essex, while the South Eastern group contained

south-west Essex (1968:4-5). However, Wright notes that, even though he had extensive

material, “it is often insufficient to enable me to give the exact geographical area over

which many grammatical phenomena extend” (1968:1), and thus he states that the

boundaries and classification of the dialects is arbitrary. Taking this into account, the

grammar is organised in two ways: variant lists according to Old English phonemes and

50

variants listed according to the individual words. However, with respect to the actual

description of the dialectal variants demonstrated by informants, Wright does admit that

he did not set out to present vowel systems in such minute detail as Ellis, since, he claims,

“this grammar would have been of little use to students of English philology” (Wright

1968:11).

2.4.3

Kurath and Lowman (1970)

The focus of this study was only across Southern England and descriptions do not reach

dialect areas north of Warwickshire to the west of England and Lincolnshire to the east.

A specially designed questionnaire was used to investigate approximately three locations

within each of the southern counties. The questionnaire Kurath and Lowman used held

nearly 250 items, which were intended to highlight regional differences in phonology, and

they used the Middle English phonemic system as their frame of analytic reference when

trying to determine the normal development of sounds from the parent phoneme. The

corpus of 56 speakers was collected throughout the 1930s and this Kurath and Lowman

note “appears to suffice for a general view of the dialect structure of this section of

England” (1970:1).

Once the data was collected and analysed, they produced maps highlighting the reflexes

or variants of Middle English vowels used by the informants of each location. Thus, the

type of variants elicited through the questionnaires are clearly demonstrated across

regions, although details of specific places where the data collections took place are not

provided.

51

2.4.4

The Survey of English Dialects (SED)

The ultimate aim of the SED, conducted between 1950 and 1961, is described by Orton as

being “the compilation of a linguistic atlas of England” (Orton 1962:14). The

methodology involved the completion of an extensive questionnaire (administered by a

variety of third parties) by mostly rural male informants over the age of sixty, since these

informants were regarded as being more reliable sources of vernacular speech (Orton

1962:15). Included in the survey was a single speaker from East Mersea, as well as a

speaker from the nearby villages of Tiptree and West Bergholt, highlighted on the map

below. Therefore, it will be these three data sources which will be used to give specific

information relating to the linguistic variation appropriate to the location under study in

this thesis.

2.1 From http://www.multimap.com/maps/#t=l&map=51.84169,0.86927|12|4

West Bergholt

Tiptree

Mersea Island

52

2.5

Summary

This section has described the methodology employed during data collection on Mersea

Island and its subsequent analysis. In order to extract the most appropriate data for

analysis, casual sociolinguistic interviews were conducted with two distinct age groups of

native Islanders, thus enabling an apparent time analysis of any linguistic changes. In

addition, the discovery of an audio archive held by the local museum allows for a real

time study to be conducted, helping to substantiate any claims from the apparent time

analysis. This data, unlike that of the historical dialect studies outlined above, such as

Ellis (1889), feature conversational data instead of that collected from questionnaires, for

example, and will therefore lend a different perspective to the historical data.

In order to prepare these data for statistical analysis, the coding process has been

discussed, as well as the particular distributional restrictions for each diphthong in

monomorphemes. These distributional restrictions inhibit the data in terms of following

phonological environments and so additional domains, such as syllable type and

preceding phonological environment, will also be analysed.

53

Chapter 3

The Historical Derivation and

Variation of the Variable ()

54

3. THE HISTORICAL DERIVATION AND VARIATION OF THE VARIABLE ()

The following sections will discuss the derivational history of the present-day diphthong

// from Middle English /u:/ as well as identify and discuss past and present variation

reported from studies which have examined this variable.

3.1

Historical Sources

Wells (1982a) classifies the stressed diphthong // as the MOUTH vowel. It may occur in

both open and closed syllables (such as how and hound), though its distribution is

restricted with respect to following phonological environments (see Chapter 2.3.1.2.1

above). The MOUTH lexical set represents the standard stressed diphthong // which is,

in almost all cases, the present day derivation (or reflex) of Middle English /u:/, which

diphthongised as part of the Great Vowel Shift. The phonological ‘event’ known as the

Great Vowel Shift, which began around the fifteenth century, refers to a phonological

change that took place in English, which resulted in the raising of all long vowels and the

diphthongisation of the two high vowels /i:/ and /u:/.

However, the path of development from Middle English /u:/ to present day // remains a

topic of academic debate. This is also true for the developmental path of the diphthong

/a/ which appears to have acted, for the most part, in parallel with // (the path of this

diphthong will be discussed in Chapter 5). However, with respect to //, the opposite

views focus on the relative timing of the centralisation of its nucleus. The first view,

supported by, for example, Wells (1982a:185) and Lass (1987), reports that centralisation

of the diphthong nucleus followed an initial lowering period. Thus, the route from /u:/ to

55

// was one of lower-centralise-lower. The following is an illustration of this

developmental path:

3.1 The developmental path of /a/ following the lower-centralise-lower route (based on

Yamada (1984:61))

The alternative approach suggests that the centralisation stage occurred first, before any

lowering took place (for example, Stockwell (1975)) and thus, a centralise-lower-lower

scenario is advanced.

3.2 The developmental path of /a/ following the centralise-lower-lower route (based on

Yamada (1984:61))

56

These conflicting accounts of how modern standard // developed have a significant

impact on how we may view regional dialects that exhibit variants, which are either found

on one of these proposed paths (thus representing conservatism and a standard

development), or are not realised in one of these paths at all (thus representing some

divergence from the standard).

It shall be seen that common variants of (a) in South Eastern English dialects have a

front, half-open //-type vowel as their diphthong nucleus (that is, variants not accounted

for in the models illustrated above). Therefore, with respect to the contrasting

developmental paths shown above, they do not allow for the presence of // nuclei,

suggesting that any further raising of the /a/ nucleus occurred after [a] had been reached.

Both the paths illustrated above are incorporated into Ogura’s (1987) highly detailed

representation of the development of Middle English u, which was based on calculations

from data included in the Survey of English Dialects (SED). Also included in her model

is the later development of /a/-nucleus fronting. She notes that, across the 311 locations

that she examined, there were 45 reflexes of Middle English /u:/ in England stating that

“the first element of the diphthong may be almost any vowel” (1987:40).

57

3.3 The Stages in the Development of Middle English u. Note: the main developmental

routes are circled (from Ogura 1978:41)

However, even though the respective models differ in the developmental path of //,

they appear inadequate with respect to accounting for diphthongal qualities which are not

found along the standard dialect’s phonological path – for example, the half-open fronted

[] ~ [e] nuclei which are found in many southern dialects of English in Great Britain.

For these models, the development of [] ~ [e] nuclei is a result of further fronting and

raising relative to the standard [a] nucleus. Indeed, not even Ogura’s highly detailed

model allows for the development of [], for example, until [] has been reached. The

exception is her ‘Type 2’ development which sees, from the initial centralisation to [u], a

fronting towards [] before a lowering and backing to [a], but this, she suggests, relates

only to the south-western part of England (and thus not, for example, in East Anglia

where [] nuclei have also been found). Wells (1982a:256-257; 1982b:306-310) suggests

that the fronting and raising of the MOUTH vowel in some southern English varieties (as

well as those in the Midlands (UK), Australia and New Zealand) is part of a chain shift

58

labelled as the Diphthong Shift, which also affected the long vowels /i: u:/ and the

diphthongs /e a o/. This shift is thought to have become widespread in the first half

of the nineteenth century, at the same time that British English was being transported to

Australia (Wells 1982a:257). A simplified version of the proposed vocalic movements of

the Diphthong Shift can be illustrated as follows:

3.4 The path of the diphthong shift in both front and back vowel systems (from Wells

1982a :256)

Even though it is not marked on Wells’ diagram above, which only shows a fronting of

the /a/ nucleus, he does note that this vowel may also raise to [] after the initial

fronting to [] (Wells 1982a:257)

Labov accounts for the change from [a] to [] within the model of the Southern Shift

(1994:208) – a similarly formulated set of changes to that of the Diphthong Shift. This

incorporates changes to a number of vocalic elements across four sub-sets of movement

and includes the nuclei fronting of back-gliding diphthongs (such as // and //). As a

result, the diphthong nucleus of /a/ is fronted and raised to the region of /e/ (while the /e/

59

nucleus of /e/ is lowered towards the /a/ region). Once again, the appearance of //-type

diphthongs is treated as a continuation of developments from the [a] position.

In contrast, Britain (2008) presents a range of evidence (including historical, geolinguistic

and dialectological descriptions) which dispute the above approaches to the development

of the // nucleus with respect to varieties of Southern and Eastern England, Australia,

New Zealand, the Falkland Islands and parts of the Southern United States. He argues

that, rather than these dialects being highly innovative with respect to the subsequent

development of the [] nucleus from the [a] nucleus, there is no evidence to suggest that

the varieties in question had ever featured [] as a vernacular variant at the relevant

periods of development. Instead, Britain suggests that fronting occurred much earlier so

that a centralised [u] type diphthong fronted towards [u] before lowering took place to

[] (see figure 3.5 below). However, the latter development to [] is only applicable to

British English dialects.

3.5 The alternative development of () - Based on Britain (2008)

Britain notes that the establishment of this route is important as it “reemphasises the

distinctiveness and independence of non-standard varieties, even in the South of England,

from their standard counterparts” (2008:69).

60

3.2

Variation of //

As previously mentioned, Ogura remarks that the nucleus of () could be almost any

vowel and this observation is echoed by Wakelin (1977:8). In addition, Wells

(1982a:152) outlines four main types of variation that may influence the pronunciation of

():

I. “The degree of advancement of the starting point”. For example, he notes that

very fronted starting points typically [~~] are characteristic of traditional Southern

British Englishes.

– These fronted variants have been found across the south in Milton Keynes and

Reading, for example, with respect to working class speakers and speakers

representing more traditional local variants (Cheshire et al (1999), Kerswill and

Williams (2000), for example). Fronting to an // nucleus was also found in the

Fens by Britain (1991) and by Przedlacka (2001) across the four counties of

Buckinghamshire, Essex, Kent and Surrey, while investigating the presence and

nature of Estuary English in 16 year old school children.

Further afield in the UK, Mathisen (1999) observed the use of // and // nuclei in

the speech of her Sandwell (West Midlands) informants, particularly throughout

the speech of working class males. In addition, she notes that the area’s

diphthongised GOOSE vowel ([u]) may lead to pairs such as shooting and shouting

becoming homophonous.

61

II. “The degree of openness of the starting point”. For example, Wells suggests that

half-open starting points, or those even closer, of the qualities [~~] are associated

with older and rural speech (though he does not indicate particular geographic regions

for this observation).

– However, Williams and Kerswill (1999) note a pronunciation of [] is present in

their Hull data, but they fail to comment on this and limit their observation only to

the tendency for strong labialisation of the diphthong offset.

Diphthongs with a nucleus quality produced in the region of // have been

reported in other parts of the UK, however. Mees and Collins (1999) describe the

quality of /a/ in their Cardiff data as a diphthong whose starting point “is low

back rather than truly central” (1999:191), and therefore they represent it as [u]

as opposed to [u] (see also Collins and Mees (1990)). In addition, Stuart-Smith

(1999), when giving a summary of Glasgow vowels, notes that while the Glasgow

Vernacular prefers a short [] vowel for this diphthong, Standard Glasgow English

is represented by a diphthong of an [] quality. This Standard Glasgow variant is

also found in data from Edinburgh English (Chirrey 1999)

III. “The quality of the second element”. By way of illustration, Wells suggests that

an []-type diphthong featuring a front unrounded second element is represented in

some areas of Southern England. Also, a mid-central type element resulting in the

centring []-type diphthong can be found in London Cockney English. It is also

possible for there to be no second element, and thus, through a process of

monophthongisation, renderings such as [: ~ a:] are also found in London Cockney.

– With respect to the qualities of the /a/ offglide, unrounding to [] has been

reported in areas such as Reading (Williams and Kerswill 1999) and Milton

Keynes (Kerswill. and Williams. 2000). In these two locations, this form is

associated particularly with the older working class speakers. However,

62

Przedlacka (2001) found offglide lowering to [] in the data of 16 year old

speakers across the counties of Buckinghamshire, Essex, Kent and Surrey. In

contrast, the centralisation of the /a/ offglide to schwa was reported in London

Cockney (Sivertsen (1960)) and, indeed, Matthews (1938 [1972] cited in Ryfa

(fc)) states that Cockneys, who are considered less broad, employ [] as the

offglide to /a/.

Regarding monophthongisation of /a/, both Sivertsen (1960) and Tollfree (1999)

report on the use of a monophthong in London speakers. However, it can also be

found outside London, indicating that it is not a uniquely defining feature of the

city or indeed the southeast. For example, Britain (1991) notes monophthongal

forms of [e: ~ : ~ a: ~ :] qualities in his East Anglian Fens data while Docherty

and Foulkes (1999) and Stoddart et al (1999) note that [a:] is the main local

variant in their Derby and Sheffield data, respectively.

IV. “The ‘speed’ of the diphthong”, which tends to affect the articulation of the initial

element causing a prolonged articulation of the onglide (which is a particular

characteristic of the American South).

– However, others, such as Tollfree (1999), have noted diphthongs, such as [:],

where the second element has a very short duration and, thus, the first element is

prolonged.

Indeed, these are similar points of variation associated with the // diphthong (see 5.2

below) and, also in a similar fashion to //, the MOUTH vowel is susceptible to smoothing

processes, particularly in the context of // triphthong sequences. Thus, tower /t/

63

may become [ta:] or even [t:] through smoothing and subsequent monophthongisation,

leading it to become homophonous with tar (Wells 1982a:238-239).

One final aspect of () variation, which is also a shared characteristic with //, is that of

‘Canadian’ Raising (see for example, Chambers (1989) or Britain (1995; 1997)). Unlike

the variation reported for //, there have been no reports of Canadian Raising patterns in

England, and Trudgill notes that “we nowhere find different allophones of /au/ in voiced

and voiceless environments” (1986:156). The Canadian Raising process is where the

diphthong variants of // and // are in complementary distribution – those with a raised

nucleus appearing before voiceless consonants and those with lowered nuclei surfacing

before voiced consonants and all other environments. Thus, with respect to //, pairs

such as [] ~ [] versus [ld] occur.

The following sections will present the findings from a range of historical and modern

dialect studies relating to the behaviour of the () variable. The initial discussion will be

primarily concerned with data from East Anglia in general before attention is turned

briefly to some international varieties of English. This will construct a picture of what

variants are represented in the regions surrounding Mersea Island and, therefore, those

varieties with which Islanders are more likely to share sustained contact. In addition, by

looking at the behaviour of () in international insular varieties, we can examine patterns

of variation across a range of sociolinguistic environments which might share similar

patterns of contact with outside varieties.

64

3.2.1

Historical Variation

In order to establish the historical development of the () variable in the area under

analysis, this section will examine its variation as presented in a selection of historical

dialect studies. The inclusion of historical data will also enable a real time perspective of

variation and change to be established and, through a comparison with the modern data

collected for this study, the trajectory of changes relating to this diphthong may be

uncovered.

3.2.1.1

Ellis (1889)

The dialect survey conducted in Essex is described by Ellis as failing to obtain and “bring

out the chief peculiarities” (1889:221) of the region. However, the data acquired with

respect to // suggests the variation at the time included variants such as [] and [],

but Ellis notes that [] should be taken as the general sound for the county. However,

one additional observation to be made comes from data from Maldon (a town

approximately 15 miles from Mersea Island), where it is noted that an occasional prefix

of // is added to the diphthong leading to variations such as {nE’u}– [n]11

.

3.2.1.2

Wright (1905, reprinted in 1968)

The result of Wright’s dialect survey is the observation that, with respect to Old English

long u (or u ), [eu] is common in Eastern England for words such as about, house and

11

Note: the transcription using the notation of Ellis is included within { } brackets while the translated

modern IPA symbols (using Eustace 1969 for the translation) are enclosed within the conventional [ ]

square brackets.

65

south, even though [u] is present in Norfolk. In addition, OE u r is noted to have

achieved breaking through schwa insertion in the items our and shower and has the

pronunciation [u] in north-east Norfolk and east Suffolk (1968:151).

3.2.1.3


Kurath and Lowman noted that, in the central southern counties they investigated (Essex

was classified as an east-central county), Standard British English [u] did not emerge in

the speech of those under investigation whereas [u ~ ~ u] were universal (1970:5).

In addition, an isogloss pertaining to /a/ variation was constructed and ran from the

western Sussex border (with Hampshire) up towards central Worcestershire as a way of

visually separating Western [u] from Eastern [u] (1970:35). However, it must be noted

that they also observed a consistent Norfolk pronunciation, which deviated from this

Eastern form, as [u] with occasional [u ~ u] (1970:5).

3.2.1.4

The Survey of English Dialects (SED)

A comparison of three locations from the SED’s data survey (East Mersea, Tiptree and

West Bergholt, see the map in 2.1 above) shows Ellis’ assessment of [] as the

predominant variant holds true across each location. By way of illustration, the tokens

about, cloud, houses and cow demonstrate the stable nature of the variant before a

voiceless and voiced consonant, a morpheme boundary and in a word-final open syllable.

Indeed, variation only becomes apparent when // is followed by //.

66

3.6 A comparison of () and () pronunciations recorded by the SED (Based on Data

extracted from Orton and Tilling (1969; 1970; 1971))

East Mersea Tiptree West Bergholt

Diphthong about

how

cows

cloud

Triphthong

ours

hour

flour

flowers

Note that the SED’s notation []

is equivalent to modern IPA []

It can be seen that the lowered, more open nucleus is preferred in each token by the East

Mersea speaker while the Tiptree and West Bergholt data contrast particularly in the

token flowers.

Additional observations which can be made from the SED are that there is no evidence of

monophthongisation or smoothing processes as well as no evidence of [] prefixing.

3.2.1.5

Summary

These historical sources suggest that the most prolific historical variants of (a) across the

south-eastern regions discussed here are those with fronted half-open or centralised

nuclei. Therefore, in light of this evidence, we can propose that these varieties were

indeed subject to early fronting before lowering to the [a] nuclei (which is demonstrated

by the studies below), since there is little or no evidence of these lowered nuclei in the

historical data.

67

3.2.2

Modern Variation Studies – The South East

This section will give a brief overview of some sociolinguistic variation studies which

have focussed upon (), or included it as part of a dialect survey. Due to the location of

the community under study, varieties from the South East of England will be primarily

considered alongside those studies whose geographical focus is on locations closest to

Mersea Island. This will demonstrate the type of variation most likely to be in contact

with that of Mersea Island and will help to generate an overall picture of regional

variation in the south east of East Anglia.

3.2.2.1

The East Anglian Fens

The Fens are a low lying area spanning four counties: parts of west Norfolk, north

Cambridgeshire, southeast Lincolnshire and northwest Suffolk. Historically, the southern

two thirds of the Fenland area was made up of undrained marshland which was prone to

flooding, making the land unsuitable for settlement (Britain 1997:19). As a result, the

population was confined to a number of smaller settlements and communities which

scattered about the region. The reclamation of the land, as it was drained in the

seventeenth century, was a turning point in the region’s history as these once isolated

communities were now accessible. This is reflected in the extent of population growth

between 1563 and 1801 in areas such as Wisbech and March, which were inaccessible

before reclamation. The population of these areas grew from 1,210 and 1,025, to 4,710

and 2,514, respectively, and Britain notes that the origins of the in-coming population

were mixed and varied (Britain 1997:19-20). As a result, these once isolated

68

communities found themselves in high contact situations, particularly between those

dialect varieties of the eastern and western Fens.

As we shall see below, Britain (1991) found prominent Canadian Raising patterns with

respect to the () variable (see 5.4.2 in particular). However, no such pattern, even in the

historical data of the SED pertaining to the region under investigation, could be found for

(). Even though Canadian Raising variation was not found to occur with this variable,

the Fens data demonstrate certain phonological influences on the variation pattern which

was analysed.

Using data from the SED, Britain established that traditional realisations of () across

available Fenland locations were diphthongs whose nucleus was predominantly of an []-

type quality. However, Britain’s data highlighted a number of different monophthongal

variations with qualities representing [e : ~ : ~ a: ~ :]. Looking at both word-medial

and word-final environments, as well as whether // was part of a triphthong, Britain

found that following phonological environment correlated with the production of

diphthongal or monophthongal variants. For example, even though the analysis

highlighted greater diphthong usage in Eastern areas and greater monophthong usage in

Western areas, overall, monophthongs were more common when () preceded [] and

[t]. Conversely, diphthongal variants were more likely in word-final position preceding a

vowel or a pause. Building on this analysis, Britain investigated the relationship between

monophthongisation of /a/ and its onset quality by dividing his 81 informants into five

groups, according to the extent of monophthongal use and the quality of their

monophthong and diphthong realisations. However, he did not report any link between,

for example, onset quality and frequency of diphthong pronunciations (i.e. Group Five

69

exhibited almost categorical [a] onsets and diphthongal variants while Group Four,

having mostly [e] onsets, also used mainly diphthongal forms (1991:128-129)).

3.2.2.2

Colchester

In Colchester, Britain’s oldest recorded town, which is approximately a twenty minute

drive from Mersea Island and can be described as its nearest urban centre, Ryfa (2005)

explored many linguistic variables including the MOUTH vowel. Ryfa’s work focussed

upon data from two youth groups, which she categorised as ‘Chavs’ (working class) and

‘Grungers’ (middle class) – however, see Ryfa (2005) for a more detailed description of

these groups. Her study analysed the speech of eight informants, four from each social

group, and consisted of spontaneous interviews across Colchester town centre where Ryfa

approached the individuals and asked permission to record them at that time.

She found that, whereas the Grungers were more likely to use variants which

approximated standard //, the Chavs preferred more fronted diphthongs. Another

observation which can be taken from her analysis is that both groups used a

monophthongal [a:] variant (Grungers 31% and Chavs 25%, Ryfa (2005:52)), but other

monophthongs can be found, ranging in quality from a front open [:] to a back

unrounded [:]. Also, there is no case of half open []-type onglides to the diphthongs

within her Colchester data, suggesting that Colchester youths, or at least those

interviewed for this study, are utilising only lowered diphthongal nuclei for this variable.

70

3.2.2.3

Basildon

Basildon was built in the 1950s as a post-war ‘New Town’. Fox (2000) notes that the

majority of the establishing population consisted of families of working class origin who

moved out of the London’s East End districts. Therefore, the incoming London variety

would have dominated not only the other input dialects, but also the original dialect

spoken in Basildon (which was, before the new town development quite rural and

sparsely populated). Therefore, the focus of Fox’s survey of this area was to investigate

whether this original Cockney influence still remained in the Basildon dialect, or whether

the dialect had succumbed to what has been referred to as ‘Estuary English’ (a variety

which, it has been proposed, has resulted from Londoners modifying their Cockney

speech towards that of the areas to which they move and RP speakers accommodating

their dialect towards that of the incomers (Coggle 1993 cited in Fox 2000:1)).

The survey of Basildon speech utilised a rapid anonymous survey (which involved the

reading of a short passage) as well as interviews carried out at a local youth centre. All

informants were adolescents between 12 and 19 years old – 22 informants being recorded

during the interviews and 30 informants used for the reading passage survey.

This study looked at a number of phonological variables, including the MOUTH vowel,

with the aim of assessing whether the vernacular speech of Basildon “is simply ‘Cockney

gone east’ as opposed to so-called Estuary English” (Fox 2000:1). Citing a range of other

previous research, Fox describes the traditional Cockney diphthong as one with an []

nucleus, while the Cockney monophthong has an [a] quality. Conversely, Estuary

71

English, not employing the monophthong, has been shown to have diphthong nuclei of [e

~ a ~ ] qualities (Fox 2000:5-6).

With respect to the actual analysis, Fox’s data only consisted of tokens extracted from six

teens aged 13 to 15 years, the rest of the informants being excluded for reasons of non-

eligibly and low quantities of data. With respect to //, she found that the monophthong

[a:] (often said to be a defining feature of Cockney) was quite prevalent in the data and

also that the vowel has a more central offglide, lending the diphthong a [] quality. This

observation is in line with Altendorf (2003), who notes that a monophthong realisation for

() is characteristic of Cockney as opposed to Estuary English and that her data

demonstrate that it is “confined to the working class” (2003:103). This, she claims, helps

strengthen the case that Basildon English is a diffused Cockney dialect as opposed to

Estuary English.

3.2.2.4

Summary

These studies, once again, show that traditional variants in the South East, with the

exception of Cockney English and areas of the Fens, were diphthongal in nature. The

nucleus qualities of these diphthongs range between front, half open and open vowels.

Also, the distribution of these diphthongal and any monophthongal variants may be

influenced by the social factor of class as well as phonological context.

72

3.2.3

Contemporary Studies of Insular Varieties

This section will specifically focus upon studies of insular communities. These studies

will highlight how a changing demographic and socio-economic status can influence not

only linguistic variation, but also the attitudes towards that variation. The locations of

these studies will include analyses from the United States, the Falkland Islands, St Helena

and Tristan da Cunha. Through the incorporation of these studies, it will be possible to

highlight not only possible linguistic constraints that must be considered in the analysis of

Mersea Island English, but also any social and attitudinal factors which might also be

relevant.

3.2.3.1

Martha’s Vineyard

Martha’s Vineyard is an island off the coast of Boston, Massachusetts. This was the

location of Labov’s groundbreaking sociolinguistic study (1963 reprinted in Labov

(1972)), which investigated the correlation between the variants of () and ()12

and the

many social factors pertaining to the community under observation. The particular

variation under investigation was the distribution of raised or centralised nuclei of the

diphthongs. At the time of the study, Labov notes that, with respect to data from the

Linguistic Atlas of New England, Martha’s Vineyard // shows little centralisation

whereas // is “well centralised” (1972:11).

1,500 tokens of () were extracted from 69 speakers, and the results showed that

centralisation seemed to be phonologically conditioned, with respect to following

12

Note that Labov used the notations (ay) and (aw) for () and (), respectively. The latter pair has been

retained here, and throughout this thesis for consistency.

73

environment. Thus, voiceless obstruents were found to favour centralisation while

sonorants and nasals disfavoured it (see Labov (1972:20) for the full scale). Although

Labov does not say as much, this pattern of centralisation coincides with the type of

environments found in Canadian Raising patterns.

Centralisation was also discovered to increase successively from the oldest speakers (75+)

to the 31-45 year old group, which represents the peak of centralisation across all the age

groups. The final age group, the youngest at 14-30, however, shows a decline in

centralisation (by almost half compared to that of the 31-45 year old group). Indeed,

Labov notes that, since the level of centralisation is at a minimum for both the very old

and the very young, it “shows that the effect of age cannot be discounted entirely, and it

may indeed be a secondary factor in this distribution over age levels” (1972:24).

Another significant factor was found to be the speaker’s social orientation and attitude to

the Island, in that a positive orientation towards Martha’s Vineyard strongly correlated

with higher degrees of centralisation (Labov 1972:39).

These findings were supported by Pope et al (2007) who reconstructed the original

Martha’s Vineyard study. Keeping the original methodology intact, 116 speakers were

interviewed and 2,500 tokens of () were extracted. They too found correlations with

phonological context and, particularly, their attitudinal assessment found that there

remained a correlation between speakers’ attitudes and the amount of centralisation

(2007:621). However, they noted that both () and () variables did not seem to be

salient (or above the level of consciousness), and that their data, collected in 2002,

74

showed levels of centralisation receding (with () not being affected as much as (a)).

This was particularly true with speakers born after 1979 (2007:623).

These studies indicate that the additional aspect of social orientation may be a significant

factor in shaping sociolinguistic variation. Types of attitude may act to inhibit or,

conversely, promote certain types of change depending upon a variant’s social evaluation

when introduced though contact. The type of contact experienced by Martha’s Vineyard

Islanders with the mainland also appears to be reciprocal. That is, even though many

tourists and permanent residents come to the Island, the Islanders also travel off the Island

for professional or social activities (in this respect, the contact situation on Martha’s

Vineyard is similar to that of Mersea).

3.2.3.2

Ocracoke and Smith Island

Ocracoke and Smith Island are two islands off the coast of the Eastern United States

which are only accessible by ferry. Ocracoke Island is situated off the coast of North

Carolina and has been subject to a significant rise in the level of tourism in recent

decades. However, Smith Island, located in the Chesapeake Bay area of Maryland, has

not had such a tourism boost and, instead, has even begun experiencing a decline in its

residential population (see, for example, Schilling-Estes and Wolfram (1999), Schilling-

Estes (2002) and Wolfram (2008)).

75

Schilling-Estes and Wolfram (1999) illustrate that, with respect to //13

patterning, the

following alternations may occur:

Nucleus – Raised [ ~ ] vs. Unraised [ ~ ]

– Fronted [ ~ ] vs. Unfronted []

Offglide – Fronted [i ~ I ~ ]

– Non fronted [u ~ U ~ o]

– Absent [e.g. monophthong]

From Schilling-Estes and Wolfram (1999:497)

They note that glide fronting is accompanied by a fronting of the nucleus by Ocracoke

speakers, but this is not always the case for Smith Island speakers. In addition, it is these

glide-fronted diphthongs which “feature prominently in discussions of Smith Island

speech, both by residents and outsiders, but almost never mentioned in discussions of

Ocracoke English” (1999-490), and this noticeably contrasts with the social saliency of

() (see 5.5.2 below).

However, even though it seems each location attributes different symbolic values to ()

and (), both islands were shown to have increased levels of // glide-fronting between

old and middle aged speaker groups, with middle aged females having greater levels of

glide-fronting on both islands. Yet, despite this feature being on the increase, it seems

that this fronted glide is restricted to closed syllables with both Ocracoke and Smith

Island speakers preferring back-gliding variants in word-final position (for example, [nau]

now).

Once again, these studies demonstrate the influence of social attitudes on sociolinguistic

variation. However, these studies also demonstrate how the same variable may be subject

13

Once again, // will be used instead of /aw/ and // in place of /ay/ when referring to these works in

order to maintain consistency.

76

to contrasting social evaluations in different communities. In addition, the data

demonstrate how an almost allophonic distribution of variants may be found according to

phonological environment. This was also noted for Martha’s Vineyard above and it may

suggest that, in insular communities where there is increasing contact between dialect

varieties, these types of allophonic alternations may arise as a result of phonological re-

allocation (see Chapter 9 for a more in-depth discussion of this point).

3.2.3.3

The Falkland Islands

The Falkland Islands are situated in the South Atlantic Ocean with the British laying

claim to them in 1690. Sudbury (2000), Sudbury and Britain (2000) and Britain and

Sudbury 2010) note that colonisation of the Falklands contrasts with other colonial events

across the English speaking world, in that there was no indigenous population with which

the settlers could have contact. Thus, the original colonising population (consisting of

predominantly British nationals) had little contact with other language forms14

during the

initial settling period.

Across the nineteenth and twentieth centuries, sporadic immigration from areas across

South America, Scandinavia and the British Isles resulted in a population peak of 2,400

island residents in 1931. However, many of those coming to the islands did not become

permanently established there and thus the Island population was transient and in a state

of fluctuation. Sudbury goes on to note that, even though the population and economy

declined dramatically following the Second World War, the modern population consists

of only 60% ‘native’ Islanders. The remaining population consists of 30% British

14

Although Sudbury notes that a small proportion of early settlers were from Northern Europe

77

residents with the remaining 10% being made up from a diverse mixture of Europeans,

Americans and Asians (Sudbury 2000; Sudbury and Britain 2000; Britain and Sudbury

2010).

With respect to /a/, Sudbury (2000) notes that it is unlikely that the range of early

Falkland settlers (mainly from the West Country and Scotland) would have brought

diphthongs with open nuclei and that the main variants would have been those with mid-

central and mid-front nuclei qualities.

The subsequent analysis of Falkland Island English (FIE) resulted in 4695 tokens of (a),

which were elicited through casual, unstructured sociolinguistic interviews. From these

tokens, diphthongs with onset qualities of [a ~ ~ ] were found as well as degrees of

offglide weakening15

, the latter being a characteristic of other Southern Hemisphere

Englishes. The analysis of onset quality revealed fewer diphthongs with open nuclei were

produced before voiceless consonants, suggesting a Canadian Raising type pattern.

However, Sudbury notes that “this allophonic patterning is less striking than in other

varieties” (2000:275) and highlights significant variation across the behaviour of FIE

speakers. This subtle allophonic pattern may reflect the natural direction of change from

half-open to open nuclei in this dialect. Trudgill (1986) notes that the Canadian Raising

pattern may result from a natural re-allocation of variants during dialect contact and

language change and thus, the observations made with respect to the Falklands data

resembles that of the Mersea Island analysis (See Chapter 9 for further details).

15

Sudbury (2000) distinguishes between a full [u] offglide, a weakened [V] offglide and a monophthong

[V:]

78

With respect to the quality of the offglide, Sudbury found that younger speakers were

more prone to glide weakening, indicating a change towards the weakened

monophthongal forms, and that females seemed to be leading this change away from

diphthongal forms. However, Sudbury does note that the “overall scores are very similar

between groups” (2000:277). Even though the levels of glide weakening and

monophthongisation of () are more frequent in the younger speakers, there is no

parallel behaviour in the offglide of () in Falkland Island English (see section 5.5.3

below for a description of this variable). This suggests that either () is more susceptible

to these weakening processes or that, if monophthongisation is a continuation of

developments to the diphthong, changes to () are more advanced than those relating to

().

3.2.3.4

St Helena

English on the island of St Helena, in the South Atlantic Ocean, can be traced to 1658

when it was colonised by the East India Company. Schreier (2008; 2010a) demonstrates

that the island’s population has undergone extensive language contact and ethnic mixing.

The founding community represented Europeans from England and France as well as

West Africans, Indians and Madagascans. Schreier goes on to show how this population

was also socially diverse, featuring a mixture of soldiers, planters’ servants and slaves

brought by the East India Company and that, although not much is known about the

nature of the British settlers, “many – if not most – of them came from Southern, perhaps

South-eastern England” (Schreier 2010a:225). In addition, it seems likely these British

settlers represented those from the working classes.

79

Travel to and from the island is incredibly restricted and, even though the population in

2010 was recorded as approximately 4,000 (Schreier 2010a), this is a significant

reduction from the population recorded a decade before. This is due to around 30 per

cent of the population leaving St Helena once the United Kingdom granted the islanders

full citizenship rights.

After integrating himself successfully in the community and establishing substantial

network ties, Schreier collected a corpus of 33 informants ranging across the regions of

the island. From this corpus, 20 informants born between 1911 and 1940 were selected,

equally representing both genders (Schreier 2008:162-163). Though there is no

indication of early forms of the // diphthong, the modern analysis of () resulted in the

observation of [ ~ ~ ] as the most usual pronunciations across the islanders

(Schreier 2008; 2010a). Even though there was no apparent Canadian Raising pattern (as

observed for the /a/ diphthong, see 5.5.4 below), Schreier observed a distinction between

the behaviour of some function words and lexical words. Indeed, the backed // quality

of the diphthong nucleus was noted for the function words out and about, while the

nucleus for lexical words tended to be more fronted (but not so fronted that it encroached

on the vowels // and //).

In addition, in contrast to the behaviour of Falkland Island English (see above), glide

weakening of St Helena is reported by Schreier to be uncommon in the speech of his

informants. Thus, full diphthong offglides are maintained.

The lack of contact with outsider varieties in recent history may account for the lack of

Canadian Raising type patterns for this variable. However, it may also be the case that a

80

change towards [] has been completed and the allophonic distribution was only

temporary and, thus, has been lost from the dialect (compare this with the behaviour of

() in 5.5.4 below).

3.2.3.5

Tristan da Cunha

Situated in the South Atlantic (south of St Helena and west of Cape Town, South Africa),

Schreier notes that, as of June 2008, the variety known as Tristan da Cunha English

(TdCE) is spoken by a community of only 278 speakers (Schreier 2010b:245). The

development of TdCE began in the 1820s and the initial mix incorporated many varieties

across the British Isles, Northeast United States, South Africa and St Helena and present

day speakers are monolingual with respect to the TdCE variety (though Schreier notes

that there are occasional borrowings from Afrikaans (2010b:245)).

With respect to significant historical events, in the autumn of 1961, volcanic activity on

the island forced a mass evacuation. However, these evacuees, who were mainly based in

Southampton (UK), returned home from England in 1963, but this two-year break

encouraged the modernisation and adoption of Western culture by the islanders. Social

stability has since then returned and Schreier observes that prolonged or permanent out-

migration is rare (2010b:248).

With respect to variation of //, Schreier writes that the most usual pronunciations are of

[ ~ ] qualities and that there is no evidence of Canadian Raising patterns for this

diphthong. Indeed, Schreier notes that since neither TdCE nor St Helena English (StHE)

exhibit Canadian Raising patterns for //, but do for //, this is evidence of the influence

81

that StHE had on the historical formation of TdCE (2010b:230). Thus, in a similar

fashion to St Helena, it may be the high degree of isolation which is preserving the

Canadian Raising type pattern of (), rather than allowing the change to complete as it

may have done for ().

3.3

Summary

This section has discussed the historical variation and derivation of the variable (). We

have seen that present-day // derives from Middle English /u:/, as a result of

developments associated with the Great Vowel Shift. Even though the path of //’s

development appears uncertain and controversial, historical evidence from Essex seems

to support Britain’s (2008) proposal of earlier fronting towards [] before lowering to

[] occurred.

With respect to more modern sociolinguistic studies, it has been observed that ()

variation can include a range between diphthongs which have a half-open [] nucleus

(generally those representing older, more traditional pronunciations), and a lowered [a] or

lowered and backed [] nucleus (which are generally more innovative, modern variants).

These qualities also reflect a similar range of monophthongal vowel variants. It has also

been observed, through international island studies, that the saliency of () variation can

be associated with social attitudes, which may have a direct influence on the pattern of

both intra and inter-speaker variation. We have seen that the direction of change from

some of these international studies reflect the trajectory of [] to [a] nuclei in South-

eastern British Englishes despite having quite contrasting socio-demographic histories.

Finally, these studies have also demonstrated that Canadian Raising patterns are

82

uncommon with respect to () variation, although other variation patterns, such as

variation between lexical and functional words (St Helena) and variation according to

syllable structure (Ocracoke and Smith Island), have been noted.

83

Chapter 4

() - An Analysis of Social and

Linguistic Factors

84

4. THE ANALYSIS OF () 16

The following sections will present and discuss the results generated from tokens

extracted from Mersea Island data with respect to the variable (a). The social

dimensions of age and gender will be considered before attention is turned to the

discussion of relevant linguistic factors.

4.1

() - An Analysis of Social and Linguistic Factors

Initially, tokens of both the diphthong /a/ and the triphthong /a/ were extracted from

the audio recordings. However, only 98 tokens (55 from the Older MIE speakers and 43

from the Younger MIE speakers) of /a/ were present in the data, which compared to the

2942 tokens pertaining to the diphthong. Therefore, these were omitted from the final

analysis, due to the number of tokens being deemed insufficient for a truly representative

and comparative analysis. Before this decision was made, however, all tokens were

coded according to the variant produced, as well as for the social and linguistic factors

outlined in 4.2 and 4.3 below:

Diphthongal Variants Monophthongal Variant

a a

aa

16

Examples of (a) tokens according to phonological environment can be found in Appendix D (1a) and

(1b)

85

As previously described, age was categorised according to whether the data was from the

museum archive or from sociolinguistic interviews of Older and Younger Islanders

carried out on the island of Mersea in 2006 and 2007. As the table above shows, there

was little variation in the quality of the diphthongal offglide. Therefore, due to the

complexity of the variant spread, the majority of this analysis will focus upon percentages

relating to variation regarding the diphthongal nucleus types:

- type

- type

- type

- type

a - type

aa - type

The notation ‘aa’ will be used to represent the monophthongal variant which is more

typically represented as [a:] so that it is distinguishable from diphthongs with an [a] type

nucleus. This notation will also help the examination of the monophthongal variants in

Chapter 10, since it allows for a quality to be assigned to each timing-tier slot.

4.2

Social Factors

This section will outline the results pertaining to (a) with respect to the social factors of

age and gender.

86

4.2.1

Age

The tables in (4.1a) and (4.1b) below illustrate the token numbers and corresponding

percentages for each variant of (a) according to the age of the speaker while (4.1c)

presents this data in graphic form.

4.1a

Number of () tokens according to age

Total - Tokens

Young 1 4 2 363 182 178 730

Old 18 9 705 157 361 51 366 180 49 1896

Museum 14 2 16 196 45 25 2 10 5 1 316

Total 14 20 25 901 203 390 55 739 367 228 2942

4.1b

Percentage of () variants according to age

Young 0.14 0.55 0.27 49.73 24.93 24.38

Old 0.95 0.47 37.18 8.28 19.04 2.69 19.30 9.49 2.58

Museum 4.43 0.63 5.06 62.03 14.24 7.91 0.63 3.16 1.58 0.32

87

4.1c

Overall percentage of (a) variants by age

It can be seen that, while the speakers from the museum recordings generally prefer the

traditional [] form (62%), they do produce, albeit in small amounts, raised and

centralised []

17 and [

] forms as well as more standard-like [a] forms. Conversely,

the Younger speakers are confined almost exclusively to variants with an [a]-type

nucleus, while the Older speakers demonstrate a range between []-type and [a]-type

nuclei. Indeed, since this data does not show variation in the phonetic quality of the

offglide (the only variation being in its duration, which could be attributable to extra-

linguistic factors such as the pace and rhythm of speech), the variant data above can be

recalculated according to the ‘type’ of onglide, in order to illustrate age trends more

easily.

17

It is this variant which corresponds with the [] prefixing observed in Ellis (1889), see 3.2.1.1 for details.

0

20

40

60

80

100

(N = 14)

(N = 20)

(N = 25)

(N = 901)

(N = 203)

(N = 390)

(N = 55)

(N = 739)

(N = 367)

(N = 228)

Museum Old Young

88

4.2a

Number of () tokens according diphthong type and age

Total

Museum 16 16 241 27 15 1 316

Old 18 9 862 412 546 49 1896

Young 1 6 545 178 730

Total 34 25 1104 445 1106 228 2942

4.2b

Percentage of () variants according to diphthong type and age

Museum 5.06 5.06 76.27 8.54 4.75 0.32

Old 0.95 0.47 45.46 21.73 28.80 2.58

Young 0.14 0.82 74.66 24.38

4.2c

Overall percentage of (a) variant by age

Thus, the transitional nature of the Older speakers from more traditional variants (used

mainly by the Museum speakers) on the left of the graph to more innovative variants

(used by the Younger speakers) on the right is now clearer to see. The above data also

0

20

40

60

80

100

(N = 34)

(N = 25)

(N = 1104)

(N = 445)

(N = 1106)

(N = 228)

Museum Old Young

89

demonstrate a transition from [] to []- type diphthongs in the speech of the Museum

speakers, as opposed to a shift from [] to [a]- type variants, which would be predicted by

the models outlined in figures 3.1 and 3.2 in the previous chapter.

It can be suggested that the extent of the variation demonstrated by the Older speakers

marks or highlights the generational period in which the change from []-type to [a]-type

variants was at its most prolific. Indeed, the period in which these speakers were growing

up coincides with periods of sharp demographic change (see Chapter 1 for a discussion on

the changes to the Island’s demographic structure).

4.2.2

Age and Gender

Unlike the Older and Younger speaker groups, the only suitable speakers for analysis

from the museum recordings were male. Therefore, the variant patterning of this speaker

group is akin to the data which have already been presented in both 4.1 and 4.2 above.

However, for ease of comparison, this data set is repeated below.

90

4.3

Percentage of (a) variants according to age and gender

Once again, the range of variation produced by the Older speakers can be seen to be

demonstrated by both genders. However, the data show that, while there is little

difference between the Older females’ use of traditional []-types and Standard English

[a]-types, the males clearly prefer the former. Another observation that can be made with

respect to the Older speakers’ data is that, not only do these females seem more advanced

than the males in the change from []-types to [a]-types, but they also demonstrate a

greater use of the monophthong than the males (35 tokens, 4%, as opposed to 14 tokens,

1%, respectively).

However, this is not the case when we consider the Younger generation. The graph in 4.3

illustrates how the young males, now almost completely employing [a]-type variants, are

favouring the monophthong more than the females (105 tokens, 29%, versus 73 tokens,

20%, respectively) while females are, once again, favouring more standard-like [a]-types.

0

20

40

60

80

100

Male (N = 367)

Female (N = 363)

Male (N = 1036)

Female (N = 860)

Male (N = 316)

Young Old Museum

91

This may suggest that, while [a] is being developed as a supralocal form (as opposed to

a simple assimilation to the Standard variety), the monophthong is being treated and

adopted as a vernacular variant, which is not necessarily a continuation or progression of

the [a] variant.

With respect to group affiliation, sociolinguistic research (such as Trudgill (1974)) has

shown that males will tend to show greater use of vernacular forms regarding stable

variants. However, it has also been shown that females seem “to initiate changes above

the level of social awareness [and that] such changes tend to be in the direction of prestige

norms” (Coates 2004:185). Thus, one possible interpretation of the Mersea data is that

the change from [] to [a] is above the level of social consciousness and, therefore,

females of both Younger and Older groups gravitate towards and select the [a] variant

(indeed, it could be suggested that it is the Older females who are leading this change

from [])18

. Conversely, the males of these respective generations retain the non-

standard, traditional [] variant and lag behind in the change towards [a] until it is

established within the community. Once the shift has established itself and become

stabilised, however, the males then re-orientate themselves towards another non-standard

variant with which they can affiliate (see Chapter 9 for a discussion of the variation and

change on Mersea Island with respect to all three variables under analysis).

18

See Chapter 9.3 below for a discussion of this as a change relating to dialect supralocalisation.

92

4.3

Linguistic Factors

This section will discuss and present the details of () variation according to the

linguistic factors of syllable type and phonological environment.

4.3.1

Syllable Type

As already mentioned, (a) tokens were coded according to whether the variable was

found in the nucleus of an open syllable (such as in ‘cow’ /ka/) or the nucleus of a closed

syllable (such as ‘cows’ /kaz/). The data below shows the overall results for the

complete variants of (a) across the Mersea Island data.

4.4

Overall percentage of (a) variants according to syllable type

It seems that the monophthong does not demonstrate significant bias towards open or

closed syllables. However, an interesting observation is that closed syllables correlate

0

20

40

60

80

100

(N = 14)

(N = 20)

(N = 25)

(N = 901)

(N = 203)

(N = 390)

(N = 55)

(N = 739)

(N = 367)

(N = 228)

Open Closed

93

more with shorter offglides, while open syllables correlate with fuller offglides. This can

be seen more clearly if variants are combined with respect to their offglide status (i.e.

whether the offglide was ‘full’ or ‘short’).

4.5

Overall percentage of full and short diphthongal offglides according to syllable type

Even though the differences are slight, this may indicate that the production of a shorter

offglide is linguistically constrained rather than being a by-product of language change

towards monophthongisation. Indeed, as the following data show, while variants with

fuller offglides always remain the more dominant of the pair, the extent to which they are

produced is reduced when found in the context of a closed syllable. However, the graph

in 4.6 illustrates how this behaviour is declining across the generations.

0

20

40

60

80

100

Full (N = 2064)

Short (N = 650)

aa (N = 228)

Open Closed

94

4.6

Percentage of full and short diphthongal offglides according to age and syllable type

4.3.2

Preceding Phonological Environment

Tokens were coded by the phonological environment in which they occurred. A coding

of none represents tokens which are word-initial, such as ‘out’. Once again, since there

was no qualitative change to the diphthongal offglide (that is, it remained as a high back

rounded vowel), the following data will illustrate variant types as opposed to each

individual variant.

With respect to preceding environment, the data in 4.7 presents the overall percentages of

variant types according to preceding manner of articulation.

0

20

40

60

80

100

Open (N = 206)

Closed (N = 524)

Open (N = 483)

Closed (N = 1413)

Open (N = 60)

Closed (N = 256)

Young Old Museum

Full Short aa

95

4.7

Overall percentage of (a) variants according to preceding manner of articulation

Due to low token numbers in the context of a preceding vowel and preceding /sk/ (1 and

4, respectively), these environments will not feature in the subsequent discussions

(particularly with observations relating to whether the preceding s-cluster correlates with

the properties of the second consonant - in this case whether the dorsal stop in /sk/

patterns with the dorsals in simple onsets).

The generalisations that can be extracted from this overall data set are that tokens with

preceding stops and word-initial (a) produce more []-type variants, while other

preceding environments promote more [a]-type. These observations, together with

preceding approximants promoting [a]-types above other variants, can be found across the

data for individual age groups (Appendix A.15, A.16 and A.17).

0

20

40

60

80

100

Stop (N = 1049)

Fricative (N = 581)

Nasal (N = 443)

Approximant (N = 272)

/sk/ (N = 4)

Vowel (N = 1)

None (N = 592)

96

A similar scenario is found when analysing preceding segments by place of articulation.

The graph in 4.8 once again represents the overall results for extracted tokens across the

Mersea Island data and it can be seen that each environment displays wide-ranging

variability.

4.8

Overall percentage of (a) variants according to preceding place of articulation

The data shows that, of the 34 raised -type tokens (which represents 14 tokens of []

and 20 tokens of []), 31 (91.2%) are associated with preceding coronals19

. In addition,

these 31 tokens appear to be somewhat lexically restricted, mainly to the items now, down

and town (5, 21 and 5 tokens, respectively). These were produced not only by the two

oldest speakers of the Museum data (one born in 1883 and one in 1890), but seemed to be

preserved in the speech of two males from the Older data set, one 64 year old and one 75

year old (the latter of which was the oldest male recorded during the 2006/7 data

collection).

19

Compared to the 1252 (42.6%) tokens of preceding coronal overall

0

20

40

60

80

100

Labial (N = 573)

Coronal (N = 1252)

Dorsal (N = 69)

h (N = 451)

None (N = 592)

97

Another interesting observation is that, even though it is the source of the fewest tokens,

the most outstanding environment is that of preceding dorsals, which seem to prompt

more [a]-type articulations when compared to the behaviour of the other environments,

and this trend is repeated in the Museum and Older data sets (see Appendix A.18 and

A.19). Indeed, the preceding dorsal environment is the one of least stability in the

Museum data and the only environment which favours [a]-type variants above []-type

variants in the Older speakers’ data (55.6% for [a]-type variants compared to 22.2% for

[]-type variants).

One possible motivation for this dorsal pattern is that, as the tongue retracts towards the

back of the mouth in order to make the necessary closure at the velum, it is in a low

position when it comes to making the following vowel articulation. Therefore, it would

require more time and effort on the part of the speaker to front and raise the tongue

towards the [] target area and, since there is no phonemic distinction between [] and

[a] which would require greater articulatory precision, the articulation of [a]-types is

favoured. Thus, the patterning of the data suggests that, this phonological environment

could have been the environment which helped promote and encourage the change from

[] to [a].

98

4.3.3

Following Phonological Environment

In a similar fashion to preceding environments, variants were coded according to their

following phonological environment. However, in these cases, a code of none represents

a token which is in a coda-less syllable in a word-final position, which allows for a

differentiation of words such as ‘thousand’ /a.znd/ and ‘cow’ /ka/.

Firstly, with respect to manner of articulation, 4.9 gives the overall data for the Mersea

Island speakers. The data presented is according to variant type, as there was no

qualitative difference in the offglide which could be attributable to influences of the

following segment and, unlike the case of open and closed syllables above, there was no

apparent correlation with duration of the offglide and the manner in which the following

segment was produced.

4.9

Overall percentage of (a) variants according to following manner of articulation

0

20

40

60

80

100

Stop (N = 981)

Fricative (N = 330)

Nasal (N = 922)

Approximant (N = 3)

Affricate (N = 9)

Schwa (N = 21)

Flap (N = 62)

None (N = 614)

99

It seems that, overall, only following stops, flaps and schwa correlate with higher

proportions of []-types as opposed to [a]-types (the latter data of which were made up of

the tokens ‘now-a-days’, ‘allowance’ and the surname ‘Haward(s)’). However, this

pattern is not generally reflected by the individual age groups (see Appendix A.21, A.22

and A.23).

In a similar fashion to following manner of articulation, following place of articulation

does not seem to have any significant effect on the duration of the offglide, except in the

case of following labials. It has already been noted (in 2.3.1.2.1 above) that the

distribution of (a) is heavily restricted in mono-morphemes so that it, generally, does not

occur before non-coronals. Therefore, the data set representing following labials below

(which promote [a]-type variants more than any other environment) is primarily from

complex words (such as endowment, ploughman and allowing, the latter coded as

following [w], due to the presence of the labial glide acting as a prominent hiatus

breaker). However, this category does also include 6 tokens with following /f/s. The

source of these tokens comes from th-fronted tokens of south [saf], which breaks the

requirement for only tauto-syllabic coronals.

100

4.10

Overall percentage of (a) variants according to following place of articulation

It can be seen that coronals, glottals and word-final positions appear to promote a greater

extent of variation than other environments. However, due to the distributional

restrictions, it is not surprising that it is these three linguistic environments that are

represented by the most tokens and thus are less likely to show bias in a particular

direction due to low token numbers.

4.3.3.1

Following Voice

One final dimension for investigation regarding following linguistic environments is that

of voicing. A particular linguistic phenomenon associated with variation of (a) and (a)

(for analysis of the latter, see Chapter 6 below) is that known as Canadian Raising (see,

for example, Chambers (1989), Britain (1995; 1997) and Wells (1982c)). Canadian

Raising occurs as a linguistic rule and sees the diphthongal nuclei of /au/ and /ai/ surface

as a half-open, mid vowel when it is followed by a voiceless consonant in the same

0

20

40

60

80

100

Labial (N = 10)

Coronal (N = 1522)

Glottal (N = 775)

Schwa (N = 21)

None (N = 614)

101

syllable. Thus, pairs of words such as out [t]~ loud [ld] and house [hs] ~ houses

[hzz] (Wells 1982c:494) are derived.

The data in 4.11 present the overall percentages for Mersea Island speech. The pattern

which emerges, albeit slight, illustrates that more raised variants were produced in the

context of preceding voiceless consonants and a greater degree of lowered variants were

produced before voiced consonants. With respect to word-final (a), the Canadian

Raising literature suggests that this should pattern like those tokens with a following

voiced consonant. The separation of these three contexts shows that the expected pattern

is found in the Mersea speakers’ overall data.

4.11

Overall percentage of (a) variants according to the voicing of the following segment

This scenario is replicated by the Older speakers’ data (4.13), in which the data show a

slight correlation between []-types preceding voiceless consonants and [a]-types

preceding voiced consonants, while the Museum data in (4.12) shows that, even though

0

20

40

60

80

100

(N = 34)

(N = 25)

(N = 1104)

(N = 445)

(N = 1106)

(N = 228)

Voiced Voiceless None

102

[]-types produce the most tokens, lowered [a]-types are greater before voiced

consonants. However, these patterns are lost in the data from Younger speakers (4.14)

since there are only 7 tokens of raised [] or slightly raised [] types.

103

4.12

Percentage of (a) variants according to following voice by Museum speakers

4.13

Percentage of (a) variants according to following voice by Older speakers

4.14

Percentage of (a) variants according to following voice by Younger speakers

0

20

40

60

80

100

(N = 16)

(N = 16)

(N = 241)

(N = 27)

(N = 15)

(N = 1)


0

20

40

60

80

100

(N = 18)

(N = 9)

(N = 875)

(N = 398)

(N = 547)

(N = 49)


0

20

40

60

80

100

(N = 0)

(N = 0)

(N = 1)

(N = 6)

(N = 545)

(N = 178)


104

It was noted previously in the analysis of (a) that, with respect to following manner of

articulation, nasals appear to promote [a]-type variants and that flaps display the greatest

range of variants and correlate with monophthongisation, particularly in Younger

speakers. Both of these environments share the feature [+VOICE] which is consistent with

the finding that following voiced consonants attract [a]-type variation. Thus, we can

propose that, whereas preceding dorsals appear to lead the change from [] to [a], with

respect to following environments, it is voiced consonants (in particular, nasals) which

recognise or promote [a] and following flaps which are leading in the subsequent change

to the monophthong.

4.4

Summary

This section has presented results from the analysis of () in the data from Mersea Island

English. It has shown that there has been a change from the more traditional []-type

diphthongs towards more open [a]-type diphthongs. This transition is illustrated nicely

by the spread and use of variants (particularly the intermediate []-type diphthongs) by

the Older speakers. However, regarding the Older speakers, it seems that it is the Older

males who are retaining the traditional variants to a greater extent than females (this is in

keeping with the results for (a) but contrast with the behaviour of this group regarding

the () variable, see Chapter 8 below). With respect to the Younger speakers, the males

exhibit a higher percentage of monophthongal variants, but the overall level of this

variant is relatively small, indicating that it has not yet taken a hold within the speech

community.

105

The analysis of linguistic factors shows that a full diphthongal offglide correlates with

open syllables while shorter offglides are favoured more by closed syllables. In addition,

with respect to phonological environments, it has been seen that word-initial (a),

preceding stops and preceding labials are the highest retainers of traditional []-type

diphthongs. Conversely, preceding dorsals have a greater correlation with the lower, more

open [a]-type diphthongs and therefore may be one of the phonological environments

leading the change to [a]

Finally, an analysis of following voice has shown that a pattern similar to that of

Canadian Raising can be found in the Mersea data. Thus, raised diphthongal nuclei had a

higher percentage of occurrence when preceding voiceless consonants and lower, more

open nuclei were found to correlate more with all other environments. However, even

though such a pattern is evident in the data relating to the Museum and Older speakers,

this pattern no longer exists in the speech of the Younger informants. Instead, we can see

that it is following voiceless consonants which cause more monophthongs to surface

while diphthongs have a slightly stronger correlation with all other environments.

106

Chapter 5


Variation of the Variable (a)

107

5.THE HISTORICAL DERIVATION AND VARIATION OF THE VARIABLE (a)

This section will examine and discuss the historical derivations of the diphthong // by

initially identifying its source phoneme in Middle English before examining historical

dialect evidence relating specifically to the dialect varieties of Essex and East Anglia.

5.1

Historical Sources

The variable (), which occurs in both open and closed syllables, is also known, through

Wells’ (1982a) classification system, as the PRICE lexical set. It relates to the stressed

standard English // diphthong and has been, for the most part, historically derived from

Middle English /i:/ which, as a result of the Great Vowel Shift, diphthongised. However,

as with (), the direction of the developmental path from Middle English /i:/ to present

day // is the subject of academic debate and remains controversial. For example, Lass

(1987) supports the idea that, after the Great Vowel Shift, the nucleus of the newly

diphthongised /i:/ initially lowered before being centralised. Thus, with respect to the

standard dialect, the sequence of lower-centralise-lower process results in the present day

// pronunciation, while, it is argued, some dialects then backed the nucleus further

towards // or //.

108

5.1 The developmental path of /a/ following the lower-centralise-lower route (based on

Yamada (1984:61))

Conversely, it is also proposed that, after the initial diphthongisation process, the nucleus

first moved to a central position before lowering took place. This centralise-lower-lower

development is supported by, for example, Strang (1970) and Stockwell (1975), but

Yamada (1984:61, citing Anderson and Jones 1977) notes that both possibilities are

empirically supportable (however, see Ogura (1987) or Britain (1991) for a critical

evaluation of these contrasting approaches).

5.2 The developmental path of /a/ following the centralise-lower-lower route (based on

Yamada (1984:61))

109

However, it seems that, even though the route from /i:/ to // is different, the centralise-

lower-lower model still suggests that any backing towards //~// in non-standard

dialects must have occurred after // was reached, suggesting that dialects with backed or

backed and raised variants were more advanced in terms of the developmental chain, and

highly innovative compared to those which did not back the nucleus. Indeed, Stockwell

refers to these types of dialect as advanced dialects, as they have continued the chain of

development (1975:347).

Both of these approaches are incorporated into Ogura’s (1987) model which is based on

data from the SED. Ogura discovered 17 reflexes of Middle English /i:/ across the 311

locations from which she extracted data and these are linked according to the diagram

below:

5.3 The stages in Development of Middle English (from Ogura 1987:36)

One aspect that this model does have in common with those above is that backing

occurred only after an [] variant had been achieved. This is a similar observation to that

previously made with respect to () (see 3.1 above). An account of these developments

110

of backing and raising is provided by the Diphthong Shift outlined by Wells (1982a:256-

257) and, alternatively, Labov’s model of the Southern Shift (1994:208).

The processes outlined by the Diphthong Shift see the nucleus of the PRICE vowel back

and raise towards [], although Wells notes that raising may not always occur, resulting

in diphthongs of [] ~ [] qualities (1982a:257). Once the PRICE nucleus has raised to

the region of [], the CHOICE diphthong then raises to avoid a merger, resulting in

qualities such as []. Labov incorporates this in his model of the Southern Shift and he

claims that “there can be no question that the movements of [] and [] form a chain

shift and that the nuclei are rising” (1994:170).

However, Britain (2002b) observes that data from Ellis (1889) show diphthongs with

back mid-open [ ~ ] and central onsets were prolific in the Essex area. Therefore,

given that these variants were present in the nineteenth century, while the presence of [a]

forms was less frequent, we might want to suggest that the backing of the /a/ nucleus

(similar to the fronting of the /a/ nucleus discussed above) occurred a lot sooner in the

Essex dialect than the models above allow. The resulting path may, therefore, be

illustrated as follows:

111

5.4 The possible developmental path of /a/ in the Essex dialect

5.2

Variation of //

In a similar fashion to the variation of /au/ (see 3.2 above), Wells (1982a:149) observes

that there are four main types of variation with respect to the elements of this diphthong:

I. “The degree of advancement of the starting point”. For example, very back [] ~

[] nuclei “are characteristic of the urban south of England”.

– For example, in Derby, Docherty and Foulkes (1999) found standard [] was the

preferred variant for middle class speakers. However, this variant accounted for

only 30% of the working class data. Instead, the working class speakers in their

data (in particular, the working class males) produced a diphthong with a backed

[] or [] onglide. This male preference for backed [] onglides, with females

preferring more fronted [a] onglides, was also found by Stoddart et al (1999) in

their Sheffield data.

With respect to more southern varieties, data from Milton Keynes and Reading

show that [] and [] nuclei are only preserved in the oldest of speakers in the

former, while, in the latter, they are still maintained to a certain extent in the

112

speech of the younger informants (Cheshire et al (1999), Kerswill and Williams

(2000) and Williams and Kerswill (1999)).

II. “The degree of openness of the starting point”. For example, nuclei of a quality

similar to []~[]~[] are considered typical of the rural south of England.

– However, Przedlacka (2001) found [] nuclei for this diphthong surface in her

data for young Received Pronunciation (RP) speakers who attended Eton College.

Further afield, /a/ diphthongs with nucleic qualities of [] and [] have been

found in Cardiff English (Collins and Mees (1990) and Mees and Collins (1999))

and Glasgow English (Stuart-Smith (1999)).

III. “The quality of the second element”. For example, the offglide may be of an [] or

[] quality as a result of reduction caused by a longer nucleus.

– In her detailed discussion of Englishes of the Southeast, Ryfa (fc) does not

highlight any account of an /a/ diphthong surfacing with offglide qualities other

than [].

However, variation in the offglide quality may come about through glide

weakening in some varieties of English. For example, Schreier (2010b) notes that

the offglide quality of // in Tristan da Cunha English may surface as [] in

certain contexts due to glide weakening processes.

IV. “The ‘speed’ of the diphthong”. For example, ‘slow’ diphthongs with reduced

offglides or those in which monophthongisation occurs are characteristic of the American

South but may also be found in London.

– Docherty and Foulkes (1999) and Stoddart et al (1999) report diphthongs with

long nuclei leading to offglide reduction in Derby and Sheffield, respectively.

Docherty and Foulkes note that it is almost exclusively the working class females

113

who utilise [a::] while Stoddart et al show that if [a:

] surfaces, it does so for the

pronoun I.

With respect to the monophthongisation of //, these have been reported across

the South East England with qualities ranging from backed [:] and [:] in London

(Tollfree 1999) and the Fens (Britain 1991) to a more fronted [a:] in Colchester

(Ryfa 2005). Even though monophthongisation is generally associated with

working class speech, Kerswill et al (2006) show how monophthongal realisations

of // also correlate with non-Anglo males in inner London (as opposed to Anglos

in outer London or areas on the London periphery).

Wells (1982a) also notes that // is particularly sensitive to smoothing processes through

which monophthongisation takes place by adopting the phonetic characteristics of the

nucleus. This is followed by a lengthening process which creates a monophthong of that

quality. However, he particularly uses examples of the // triphthong to demonstrate

this process (such as where fire /f/ becomes [f]) and not the diphthong itself since

smoothing is only permitted in the context of a following vowel.

Other factors which can affect the realisation of this diphthong include following

phonological environment. The process of Canadian Raising, which has already been

discussed in 3.2 above with respect to /au/, sees the nucleus of // surfacing as [], for

example, in the context of a following voiceless consonant. Therefore, pairs such as tight

/tt/ and tide /td/ are created. Although it has been claimed by some that Canadian

114

Raising patterns do not occur in the United Kingdom20

(see 3.2 above), data presented by

Britain (for example, (1991; 1995; 1997)) strongly demonstrate a Canadian Raising

pattern of // in the East Anglian Fens (see below for details).

The following sections will report the findings from a range of historical and modern

dialect studies of () with particular focus upon variation in East Anglia in general before

attention is turned to data from international studies.

5.3


The following section will focus on data from the selection of historical data sources as

outlined in Chapter 2.4 above.

5.3.1

Ellis (1889)

With respect to the pronunciation of items representing Middle English /i:/, Ellis notes

that common pronunciations in Essex, using the translations of Eustace (1969), would be

[], [] and []. Even though there is fairly even variation across all tokens in the Essex

word list, it is worth noting the pronunciation of the tokens lice, mice and height:

20

Though Trudgill (1986) reports that different allophones of /ai/ may be found in a part of east Yorkshire,

he also notes that the phonetic forms in the alternations bear no resemblance to those of Canadian Raising

(1986:156)

115

Lexical Item Transcription in Ellis

(1889)

Vowel translation of Ellis

using Eustace (1969)

High ’i h

Kind kjnd kjnd

Wide ’ wd

Lice liis li:s

Mice miis mi:s

Height hekth hkth

Although Ellis does not comment on this variation, Ogura (1990:33) notes that Old

English y became [e:] in the Middle English period in Kent, parts of Middlesex, Sussex,

Essex and Suffolk. Later, presumably as a result of the Great Vowel Shift, this became

[i:] in present day English. Thus, some lexical items, such as mice and lice, which were

among those which adopted Middle English [e:], later developed into a long [i:] vowel

instead of the // diphthong with the rest of the PRICE set.

With respect to the pronunciation of height, the short [e] vowel has also been attested by

Wright (1968) in Essex since its origins belong to Old English e (see below for further

developments).

5.3.2

Wright (1968)

Regarding Old English , Wright notes a pronunciation of ai in dike, mine and my

(1968:127-129). However, he also notes that dry, which is derived from Old English y,

116

has a pronunciation of ‘oi’ (1968:153). As noted above, Wright observed that, in Essex, a

reflex of Old English e was the front, half open [e] vowel in the token height. He also

noted that die may be pronounced as [e] or [i]. However, there is no evidence from

either Ellis or the SED to suggest that this was an active pronunciation in the area under

investigation.

5.3.3


Kurath and Lowman’s study of Southern British English notes that the diphthong

resulting from Middle English /i:/ starts in a centralised position, leaving the most

common variant with a centralised nucleus [], though a more backed [] may also

occur. This, they note, is the case in the west of the country and East Anglia, as well as

being largely represented in the South-eastern counties

5.3.4

The Survey of English Dialects (SED) (Orton 1962)

The majority variant in the East Mersea sample was [], occurring in 52 out of a possible

74 tokens. At no point in the data from this speaker was there standard-like [] variants

present. However, other variants included:

A centralised [] e.g. dry

A centralised and raised [] e.g. bind

A more open [] e.g. tried

A lowered [] e.g. eye

and

A lowered and backed [] e.g. height

117

However, the last of these variants, [], only occurred in four tokens (height, ice, iron

and lie) and, therefore, by looking at the following environments of these words, we can

observe no potential Canadian Rising pattern. This observation also holds true for the

two nearby locations of Tiptree and West Bergholt, where [] was also the variant of

preference.

The SED data does suggest a shift since the time of Ellis’ study, however. The use of [e]

in height is no longer apparent while the long [i:] in mice and lice appears to have given

way to the PRICE diphthong. With respect to lice, both East Mersea and West Bergholt

are represented as having [], while Tiptree is variable between the long vowel [i:] and

the diphthong. Mice, on the other hand, is variable in East Mersea and Tiptree, but only

has the diphthong in West Bergholt.

5.3.5

Summary

Even though there are some instances of low back [] nuclei, these historical sources

suggest that back mid vowels such as [] and more centralised [] nuclei were the more

common variants across the south eastern counties, and Essex in particular. However,

there have been some lexical exceptions in the development of //, such as those which

have origins in Old English y and e.

5.4

Some More Recent Studies

Now that we have gained some insight into the historical development and the traditional

variants of () in the region under investigation, attention will now be turned to the status

118

of this variable, as reported by some more recent sociolinguistic studies. The initial focus

will be on those studies which are geographically close to the location under study,

Mersea Island. Thereby, an idea of the variant geography of the South-eastern region can

be ascertained. Following this will be a closer look at the behaviour of (a) in insular

communities around the world. These will help establish any patterns of variation which

may surface as a result of socio-demographic isolation.

5.4.1

Norwich

Trudgill (1974; 1988) studied the East Anglian city of Norwich (Norfolk) in both real and

apparent time through a comparison of four speech styles – word list style (WLS),

reading passage (RPS), formal style (FS) and casual style (CS). As a result of the

apparent time study, Trudgill (1974) identified the variation of Middle English /i:/ as

being representative of linguistic change. Among his arguments was the overlapping

nature of the data relating to the middle working class speakers, which featured more [

~ ] type variables compared to the data from the lower working class.

119

5.5 Variable () by class and style (from Trudgill 1974:106)

The index scores for this variable were calculated as follows:

1 = [] 2 = [] 3 = [] 4 = []

Trudgill notes that the older East Anglian form is [], while [] and [] (with tense

offglides) are RP-influenced variants, and, therefore, it may “be the case that [] is a

fairly new development in Norwich” (Trudgill 1974:106). As a result, the lower the

index score, the more standard the variants used. Conversely, a high index score

represents a closer approximation to the vernacular. Thus, the high index scores of the

middle working class demonstrate that this is the group which is promoting and adopting

the [] more than the other class groups in this Norwich data. Furthermore, an additional

argument Trudgill used to support the claim of linguistic change towards [i] was that a

higher percentage of speakers under 39 (particularly males) had this variant in their

linguistic repertoire, compared to those who were aged over 40 (1974:107).

120

A comparison of this variable in real time was made possible when Trudgill (1988)

revisited Norwich for additional data and the original methodologies were maintained,

even though he did not conduct the interviews himself. Although the quality of the

diphthongal nucleus seemed to have shifted from [], Trudgill notes that there “has also

been an increase in the use of variants of /ai/ as in nice with a back rounded onset [αi]”

(1988:40), thus confirming the apparent time findings of the previous study which

highlighted a change away from the traditional [] form.

5.4.2

The East Anglian Fens

Even though Canadian Raising patterns of /ai/ variation had been rarely attested within

the UK, Britain (1991) presented findings from the SED as well as data collected from 81

speakers located across the Fenland areas, which clearly demonstrated such a pattern.

However, he did also observe a contrasting pattern between those speakers from Eastern

areas compared with those from Western areas. He observed that, in pre-voiceless

contexts (in which Canadian Raising would promote raised diphthong nuclei), raised

nuclei are clearly a feature of Eastern dialects, while those in the West prefer more open

variants. In addition, an age comparison demonstrated that a contextual Canadian Raising

contrast of raised and open variants had retreated eastwards in favour of categorical open

variants and monophthongal forms.

By way of explanation, Britain (1991) suggested that this Canadian Raising pattern may

have come about from processes of focussing and reallocation. Thus, when the Eastern

dialects (with raised variants in all environments) came into prolonged contact with

121

Western dialects (which had open variants in all contexts) after the extensive draining of

the marshlands in the seventeenth century, it led to a focussed variety in the Central

Fenland regions. Subsequent reallocation of forms to particular phonological contexts

(namely raised variants before voiceless consonants and lowered variants elsewhere)

resulted in the type of Canadian Raising pattern that was observed (Britain 1991:96-97).

One further observation, with respect to the Fenland data, concerns monophthongisation

and the environment of (a) preceding /l/. Britain notes that females of both ages were

more advanced regarding monophthong production than their male counterparts, while

the sequence /ail/ had the greatest proportion of monophthongal forms. This suggests that

“/ail/ is undoubtedly the leading environment in the linguistic change in progress”

(Britain 1991:91).

5.4.3

Other South-Eastern Variation

In Colchester (Britain’s oldest recorded town and a 20-minute drive from Mersea Island),

Ryfa investigated the speech of two young social groups known as the Chavs and the

Grungers (see Ryfa (2005) for detailed explanation and characteristics of these social

groups). She identified eleven variants of (a) ranging from standard [a] to a raised []

and monophthongs of both raised and lowered qualities. The middle class Grungers were

found to use the standard form in 82% of tokens while the working class Chavs used it in

only 24% of tokens, preferring more raised forms instead. In addition, monophthongs

122

were found in the speech of both groups with the Grungers preferring more backed [a : ~

:] variants and the Chavs the more centralised [a:] variant (2005:50).

Other data from the South East and London areas can be found in Przedlacka (2001),

Tollfree (1999) and Fox (2007) . For example, Przedlacka’s investigation into Estuary

English shows that, in the speech of adolescents, backed and rounded [ ~ ] variants

were used in Buckinghamshire while Essex, Kent and Surrey represented [a ~ ]

variants (2001:42).

Tollfree (1999) also observes backed, slightly less rounded [] variants in her South

London data. This can be compared with her more standard London data, which exhibits

a range of more fronted [a ~ a ~ a] variants. She also notes that the second element, or

offglide of the diphthong can be centralised and is often brief or wholly lacking from

pronunciation (Tollfree 1999:168).

This variation between backed [] nuclei and more fronted [a] nuclei was also found by

Fox (2007). Her, analysis of Tower Hamlets adolescents (in the East End of London),

identified three types of backed variants [:] ~ [] ~ [] and three more fronted variants

[] ~ [a] ~ [], the latter set being regarded as the newer variants, while the backed

variants were being preserved, particularly by preceding voiced stops. This fronting, Fox

suggested, could “indicate that the PRICE vowel may be undergoing a reversal of the

diphthong shift” (2007:159). Fox also found that the ethnicity of her informants

correlated with the quality of the nucleus. For example, 93% of the Bangladeshi

speakers’ data consisted of fronted variants. In contrast, the white British and mixed

origin speakers preferred backed variants, 62% and 63% respectively (Fox 2007:160).

123

Cheshire et al (1999), studying and comparing the locations of Milton Keynes and

Reading, noted that, among the very oldest pre-New Town speakers in Milton Keynes,

there can be found raised [] and [] variants. However, in Reading, these were being

maintained by children. They suggested that the target is not the RP variant, but rather a

London-like, and perhaps supra-local, [] variant, with a low backed nucleus “which is

geographically widespread in south-eastern urban varieties” (1999:6).

5.4.4

Summary

From these studies, we can see that variation of (a) with respect to the quality of the

offglide does not tend to deviate from the high front // region. The only exception seems

to be when the offglide is weakened towards a monophthong. In contrast, the diphthongal

nucleus may be found in raised and backed positions towards the //~// regions in more

traditional dialects across the South East. This suggests that, in these dialects, lowering to

[a] had not occurred before backing and raising, which is what the diphthong shift and

the southern shift would suggest. However, backed variants, can be found in the speech

of some South-eastern adolescents suggesting that these dialect areas may be undergoing

parts of the diphthong shift, which is causing the backing of the /a/ nucleus towards [ ~

].

In addition, monophthongal realisations have been found in the speech of some younger

informants, indicating that these are innovative variants in some dialect communities.

However, as Britain (1991) noted in his Fenland data, the extent of monophthongisation

124

may be conditioned by phonological context, specifically with respect to the increased

amounts of monophthongs likely before /l/.

5.5

Contemporary Studies of Insular Varieties

As noted above, the following sections will look at selected studies from a range of

insular communities with respect to the variable (a). These studies will be able to

highlight any patterns of variation which may have been influenced by the unique social

development of each location.

5.5.1

Martha’s Vineyard

As previously mentioned, the pioneering research carried out by Labov (re-printed as part

of Labov (1972)) in Martha’s Vineyard, off the New England coast, focussed on the two

diphthongal variables (a) and (au)21

regarding the degree of centralisation of their nuclei.

With respect to (a), Labov collected 3,500 tokens and found that, even though the results

did not seem to pattern with style, a number of factors influenced the centralised variants.

Among these, phonological environment was found to correlate with the degree of

centralisation. For example, a following /t/ or /s/ was found to favour centralisation the

most while a following /m/ favoured centralisation the least These findings were also

supported in subsequent studies by Pope et al (2007) and Blake and Josey (2003) with the

latter noting that “the findings suggest that the variable presence of the Canadian Raising

rule in the speech of Vineyarders today follows the same directional pattern observed by

Labov” (Blake and Josey 2003:465).

21

Note that in Labov’s work, these were referred to as (ay) and (aw)

125

Age was also found to correlate with centralisation of (a). Labov’s apparent time data,

together with data from the Linguistic Atlas of New England (LANE), showed an

increase of centralised variants across age groups (with the exception of the very

youngest group which showed a slight decline), leading Labov to interpret this as ongoing

language change. However, these claims were only tentative as the low levels of

centralised tokens in the LANE data and the data from the youngest speakers could have

been representative of the U-Shape curve found in age-grading patterns.

However, the study by Pope et al, which replicated Labov’s original methodology as

closely as possible, did not find such a U-Shape pattern but rather a general decline in the

use of centralised nuclei. A comparison of both studies is provided below:

5.6

Centralisation of (ai) and (au) plotted separately, according to year of birth. (Labov’s

1962 study compared with the present 2002 study - From Pope et al 2007: 623)

The extent of centralisation was also observed to correlate with the attitude and social

orientation of the speakers. Indeed, as already noted above, Labov proposed that the

126

increase of centralised variants was in response to the growth of summer tourism on the

island and, thus, speakers used their linguistic resources to create and maintain their

native islander status. However, even though this was tested by Pope et al and found to

still be valid on the island, Blake and Josey (2003), who also found a decline in (a)

centralisation, reported that the iconic status of centralised /a/ had been lost and native

islanders no longer had any vested interest in trying to linguistically separate themselves

from the summer tourists. In retaliation to this, Pope et al suggested that a

misinterpretation of social data, together with clashing methodologies in the initial data

collection, resulted in these contrasting conclusions - see Pope et al (2007:624 onwards)

for a discussion of these points.

5.5.2

Ocracoke and Smith Island

As previously noted in section 3.2.3.2 above, Ocracoke and Smith Island are two

inhabited islands off the Eastern coast of the United States of America. The former is

situated off the coast of North Carolina, while the latter is located in the Chesapeake Bay

area of Maryland. Both are only accessible by ferry and while Ocracoke has witnessed a

huge rise in tourism levels over recent decades, Smith Island does not have such a vibrant

tourism industry and is even experiencing a decline in its residential population (see, for

example, Schilling-Estes and Wolfram (1999), Schilling-Estes (2002) and Wolfram

(2008)). In both islands, the diphthong /a/ is realised with a raised onset in the region of

[] or [] and, while this is a noticeable stereotype of Ocracoke English, it is rarely

perceived as a marker of Smith Island English. This is in contrast to the (au) alternations

in these communities mentioned above (Schilling-Estes and Wolfram (1999-490)).

Indeed, the salience of this feature is once again suggested as having an influence on the

127

variation patterning of each island. Therefore, where raising and backing of /a/ is salient

in Ocracoke, it is observed to be receding as a dialect feature, with the exception of the

speech of those belonging to the closely-knit Poker Game Network22

who seem to

promote and preserve traditional island features. Conversely, the 24 Smith Island

speakers showed an increase of raised and backed variants, particularly among the middle

aged and young speakers.

Another point of contrast in the behaviour of these variables comes with respect to

following phonological environment. In the Ocracoke data, the raising and/ or backing of

/a/ was favoured before voiced environments (such as in tide), which is similar to that of

the Martha’s Vineyard data mentioned above, while in the Smith Island data, raising and/

or backing was favoured before voiceless environments (for example, tight). Schilling-

Estes and Wolfram (1999:495) note that, in earlier work, this difference was attributed to

the peripheral and non-peripheral nature of [] and [

], respectively. However, in the

light of additional data from the United States, this is no longer considered by them as a

viable explanation, since this data does not fit the predicted pattern proposed by their

original model. Instead, the contrasting symbolic nature of raised and/or backed variants

is given explanatory theoretical status. Therefore, with the raised and/ or backed variant

acting, in Ocracoke, as a “symbolic icon of Islander speech and Islander identity”

(Schilling-Estes and Wolfram 1999:496), which can be used to distance locals from

tourists, only those who want to project an island affiliation will retain this feature.

However, on Smith Island, where it does not serve as a marker of island status, it is not as

readily manipulated and may be retained without socio-psychological pressure.

22

These are a group of men who regularly meet to play poker and who “project a highly (traditionally)

masculine image, and who pride themselves on speaking the authentic Ocracoke dialect” Wolfram and

Schilling-Estes (2006:239)

128

5.5.3

The Falkland Islands

The Falkland Islands are a group of islands east of Argentina in the South Atlantic Ocean.

The main resident population lives on East Falkland in the Capital Stanley (Britain and

Sudbury 2010). Even though uninterrupted Anglophone settlement can be traced back to

the 1830s (with the initial British claim going back to 1690), Britain and Sudbury

(2010:212) note that, according to the 2006 census, 55% of the population were born

outside the Falklands. This population is represented by those from Great Britain, St

Helena, Chile, Australia and a range of many other nationalities. As a result, the

Falklands “should in no way be seen as demographically homogenous or monocultural”

(Britain and Sudbury 2010:212) 23

.

With respect to the () variable, Sudbury (2000) notes that the originating population,

representing mainly Scotland and the West Country, would have brought with them

variants with [ ~ ~ ] nuclei qualities (2000:222). In addition to these, diphthongs with

open back [ ~ ] nuclei may also have been present, but those with open front [ ~ ]

nuclei “are likely to have been in the minority at the time of the Falklands colonisation”

(2000:222).

Finally, Sudbury observes that the original offglide qualities were high front [ ~ i] and,

therefore, varied with respect to the degree of tenseness. However, monophthongs were

also recorded, but these were particularly restricted to –ight lexical items (2000:222).

23

See Section 3.2.3.3 for more details of the settlement history.

129

The analysis of modern Falkland Island English (FIE) demonstrated that the preferred

variants of () had nuclei of [ ~] qualities. This suggests a degree of conservatism in

the dialect as these onsets do not seem to have progressed or diverged much from that of

the original settlers. However, Sudbury discovered a consistent Canadian Raising pattern

with respect to //. She concludes that realisations of // preceding voiceless

environments strongly favour a close mid onset, while the nucleus of // in other

environments is lower, though does not reach a fully open position. Thus, even though an

allophonic distribution akin to Canadian Raising can be found, the phonetic distinction

between allophones is less marked than in other Canadian Raising dialects (Sudbury

2000:233).

Due to the extent of settlers from Scotland, Sudbury also tested for patterns relating to the

Scottish Vowel Length Rule (SVLR). This rule generally predicts (with respect to //)

that // nuclei will be represented by open vowels only before voiced fricatives, /r/ and

word or morpheme boundaries24

. However, even though some speakers exhibited

differences in the contextual usage of () variants, Sudbury notes that these “were not

always in the direction predicted by the SVLR” (2000:243). This led Sudbury to

conclude that the SVLR does not apply as a general rule in FIE.

5.5.4

St Helena

The island of St Helena, in the mid-central South Atlantic Ocean, was recorded as having

a population of 5,644 in 1987. However, since then, due to St Helenians being granted

24

See the processes associated with Aitken’s Law in Wells (1982b) for more detailed information.

130

full British Citizenship, this population has decreased by 20.3% due to extreme out-

migration (Schreier 2008:93). Schreier (2008; 2010a) records that the founding

population of the island formed four principle groups, namely the British and European

settlers, slaves from Africa, Madagascar and Asia, and labourers from China. Schreier

goes on to write that, even though there is a lack of specific information relating to the

British Settlers, it seems likely that the London area (perhaps the Greater London area)

was the place of origin for many of them (2008:95)25

. The extensive range of island

settlers over the history of St Helena has led to a sociolinguistic environment which is

diverse and heterogeneous. However, even with such social and linguistic diversity,

Schreier observes that “its most influential founders came from the British Southeast and

Madagascar” (2008:227), while the other linguistic groups had little impact on the

resulting dialect.

Regarding variation of (), Schreier notes that [] is the most common variant (although

an open [] nucleus may also surface). However, instead of proposing that these backed

variants are preserved from older forms of the dialect, which may then lower towards [],

Schreier claims a backing of the // nucleus to the [] region has taken place, which is

indicative that St Helenian English (StHE) has undergone at least part of the diphthong

shift. He likens this to the pronunciation of Cockney English and Australian English

(2010a:234). In addition to this, the data from St Helena demonstrate evidence of

Canadian Raising. Thus, the nucleus of // was found to raise towards the mid central

region of [ ~ ] before voiceless consonants and is lowered in all other environments.

25

See Section 3.2.3.4 for additional details regarding the social history of St Helena

131

Finally, even though Schreier notes some mild glide weakening (with respect to the

second element of the diphthong, []) he does not report any use of full monophthongal

variants in StHE (2008; 2010a).

5.5.5

Tristan da Cunha

Tristan da Cunha is a sparsely populated island, with only 278 speakers of Tristan da

Cunha English (TdCE) recorded in 2008 (Schreier 2010b:245). The now monolingual

Anglophone community has developed out of a number of inputs ranging from the British

Isles to the Northeast United States to South Africa and St Helena. A significant

observation of the island community is that the population resides in one settlement

(eliminating the possibility of regional variation). As a result, the social networks of

inhabitants are very dense and multiplex in their structure. Finally, Schreier notes that the

population have a strong Tristanian identity and “their sense of orientation is focussed on

the island” (2010b:246)26

.

The modern TdCE dialect, with respect to //, exhibits an innovative allophonic pattern

between a ‘fast’ [] diphthong and a ‘slow’ [] diphthong (Schreier 2010b:249). These

allophones follow a Canadian Raising type pattern in that the former can be found before

voiceless consonants, while the latter surfaces elsewhere. The nature of the slow []

diphthong, Schreier observes, is indicative of the // diphthong having taken part in the

diphthong shift as well as glide weakening. However, there are no reports of the latter

process of glide weakening leading to monophthongisation.

26

See Section 3.2.3.5 for more information regarding the history of Tristan da Cunha

132

5.6

The Special Status of LIKE – Variation in New Zealand English

The use of like with either grammatical or discourse functions leads it to be a highly

frequent item throughout spoken language data (although different functions of like will

represent different usage frequencies). Due to the multi-dimensional nature of this item,

it is worth considering it separately from other () tokens. Even though other studies

(such as Cheshire and Fox (2007), Macaulay (2001) and Tagliamonte and Hudson

(1999)) have looked at the uses of like from the perspective of variation within the

quotative system, the following discusses how phonetic variation of like can be employed

to signal group affiliation as well as like-function.

D’Arcy observes that in “the media there is a tendency to talk of like as a single,

monolithic entity” (2007:391) and thus highlights nine categories of LIKE. These may

surface under the two broader categories of grammatical uses (for example, LIKE as a verb

or adverb) and vernacular uses (for example, LIKE as a quotative complementiser or

discourse marker). Using a subsection of these classifications (determined by frequency

of tokens and, thus, available data), Drager (2009) investigated the variation of LIKE in the

speech of girls in an all-girls high school in New Zealand. Her analysis was based on

ethnographic observations of student groups, determined by where they ate lunch. The

two primary groups were the Common Room Girls (CR) and the non-Common Room

Girls (NCR). The variation analysed included the quality of the vowel, whether the final

/k/ was realised and degrees of glottalisation.

A subsequent acoustic analysis of tokens demonstrated that the realisation of LIKE was not

only grammatically conditioned, but also socially conditioned so that speakers could

133

manipulate pronunciations as they construct their social identity or persona. For example,

in a comparison of quotative like with grammatical functions of like, her results showed

that “a token was significantly less likely to be quotative like if it was more diphthongal”

and thus, “tokens with vowels that were more diphthongal were more likely to be a

grammatical function of like” (2009:116). In addition, when comparing the realisations

of the /k/ in the quotative like and the discourse particle like, Drager observed that a

speaker was more likely to realise the /k/ in a quotative like as opposed to a discourse

particle like if she were a NCR girl and thus “more likely to drop the /k/ in the quotative

than in the discourse particle if she was a CR girl” (2009:122)27

.

5.7

Summary

Even though the historical source of present day // can be identified as mostly Middle

English /i:/, the subsequent path of derivation remains uncertain. Historical dialect

surveys suggest that a traditional variant of () in Essex is [] while in some areas,

notably across the East Anglian Fens, Canadian Raising patterns were established.

However, modern studies have shown that, in the UK, while patterns of Canadian Raising

are evident in a few locations, (a) has undergone or is undergoing linguistic change either

towards a standard-like /a/ diphthong, or towards a newly levelled or supralocal variant,

such as []. Also, it has been shown that the quality of the diphthong is considered a

salient sociolinguistic marker, not just in the UK, but in other locations such as those in

the United States and New Zealand, as presented here. In these locations, the level of

27

Drager classified a /k/ as being dropped if it could not be heard during the auditory analysis (2009:106).

If the /k/ appeared glottalised, this was marked as such but still treated as having the /k/ not realised

(2009:107).

134

social awareness regarding (a) variation may affect the variant patterning within a speech

community as speakers attempt to construct their sociolinguistic identities.

The examination of insular communities has demonstrated that the development of

allophonic distributions of () variants in a manner similar to that of Canadian Raising is

more likely to develop for this diphthong than for (). Perhaps, due to this allophonic

variation, the // diphthong is less susceptible to processes of glide weakening, in these

communities, which contrasts with the behaviour of ().

135

Chapter 6


Linguistic Factors

136


The following sections will present the results for the variable (a) in Mersea Island

English. The social dimensions of age and gender will initially be presented before a

subsequent analysis of linguistic factors.

6.1

() - An Analysis of Social and Linguistic Factors

As previously mentioned in Section 2.3.1.3, the analysis of function words (including I,

my, myself and by) will not be discussed here. Rather, this analysis will be restricted to

the behaviour of lexical words, together with a comparison of variants pertaining to the

different functions of ‘like’. Also included in this analysis, but treated separately due to

their behaviour in the data, will be an investigation into the behaviour of tokens in which

(a) immediately precedes /l/.

Following the methodology outlined in Chapter 2 above, 3620 tokens of (a) were

extracted from the data. In addition, 218 tokens of (a) preceding /l/ and 746 tokens of

LIKE were extracted. Thirteen variants were identified and these can be categorised into

the following subgroups:

1 2 3

Primary Variants Lowered and Backed

Variants

Unstressed and Reduced

Variants

28

Examples of (a) tokens according to phonological environment can be found in Appendix D (2a) and

(2b)

137

The variants in category 1 are those majority variants which are found in all phonological

environments and lexical items. Those in category 2 are more restricted in their

distribution, occurring mainly before /l/ and occasionally as a smoothed variant before a

schwa in the items ‘quiet’ and ‘lion’. The third category of unstressed variants were only

found in the lexical item ‘quite’ ([o] 4.5% and [] 2.4%) and were produced

predominantly by Younger males.

Once again, there was no significant variation in the quality of the offglide of this

diphthong. Therefore, due to the complexity of the variant spread, together with the small

overall token numbers for variants with reduced offglides compared to the full

counterparts29

, this analysis will present the percentages and analysis relating to the

following variant types:

- type - type - type - type - type - type

~ - type (fully backed and open long vowels/ monophthongs)

~ - type (unstressed)

29

The full token numbers for each individual variant can be seen in Appendix B.0

138

6.2

Social Factors

This section will outline the results pertaining to (a) with respect to the social factors of

age and gender.

6.2.1

Age

The data in 6.1 demonstrate how the variant use has changed across the three age groups.

6.1a

Percentage of (a) variants according to diphthong type and age

Young 0.00 0.00 2.67 9.24 85.88 0.84 0.31 1.07

Old 1.20 0.90 21.61 16.33 58.96 0.65 0.05 0.30

Museum 11.59 8.94 62.25 7.62 9.27 0.00 0.33 0.00

6.1b

Percentage of (a) variants according to speaker age

It can be seen that the Museum speakers, while using small levels of the standard variant,

are the greatest proponents of the -type variants, as well as featuring the greatest levels

of - type and -type variants. Conversely, the Younger speakers use small amounts of

the more rounded and - type variants and, thus, their speech is characterised

0

20

40

60

80

100

59)

(N = 45)

(N = 657)

(N = 472)

(N = 2337)

(N = 24)

(N = 6)

(N = 20)

Museum Old Young

139

predominantly by the standard-like a-type variants. It is also worth noting that, apart

from the backed and smoothed monophthongs before /l/, there seems to be no instance of

monophthongisation, such as that found with the variable (a), see Chapter 4 above. In

addition, the Older speakers, once again, show a transitional quality in their range of

variants.

6.2.2

Age and Gender

When the data is examined more closely, with respect to age and gender, we can see that

it is the males who show a greater tendency towards the more traditional -type variants.

6.2

Percentage of (a) variants according to age and gender

The Older males appear to be preserving the more traditional variants as opposed to the

females. This is also the case with respect to the Younger speakers’ data. Although their

use of -types is considerably diminished, when compared to their older counterparts, the

Younger males show similar characteristics to the Older female speakers.

0

20

40

0

60

80

100

Male (N = 640)

Female (N = 670)

Male (N = 1028)

Female (N = 980)

Male (N = 302)

Young Old Museum

140

6.3

Linguistic Factors

This section will discuss and present the details of (a) variation according to the

linguistic factors of syllable type, phonological environment and the specific lexical item

Island.

6.3.1

Syllable Type

When the overall data is analysed, open syllables are seen to correlate more with a-type

variants, while closed syllables show a greater correlation with and -types (6.3). This

pattern is replicated in the data for each individual age group (6.4), showing consistency

across the generation groups.

6.3

Overall percentage of (a) variants according to syllable type

0

20

40

60

80

100

(N = 59)

(N = 45)

(N = 657)

(N = 472)

N = 2337)

(N = 24)

(N = 6)

(N = 20)

Open Closed

141

6.4

Percentage of (a) variants according to syllable type and age group

6.3.2

Preceding phonological environment

The overall data, with respect to preceding manner of articulation, is given in 6.5 below:

6.5

Overall percentage of (a) variants according to preceding manner of articulation

0

20

40

60

80

100

Open (N = 312)

Closed (N = 998)

Open (N = 404)

Closed (N = 1604)

Open (N = 42)

Closed (N = 260)

Young Old Museum

0

20

40

60

80

100

Stop (N = 806)

Fricative (N = 561)

Nasal (N = 522)


Affricate (N = 5)

None – X (N = 85)

142

While each age environment demonstrates a similar pattern, a broader use of variants

seems to surface when (a) is word-initial (in items such as ‘eye’ and ‘Irish’) and this

pattern is borne out in each individual group (See data in Appendix B.23, B.24 and B.25).

However, with respect to preceding place of articulation, a similar pattern to that

previously demonstrated with (a) (see section 4.3.2 above) develops.

6.6

Overall percentage of (a) variants according to preceding place of articulation

Once again, dorsals are seen to promote a greater amount of lowered variants across the

overall data, and, once again, this is a trend which holds across each individual age group

(see the complete data in Appendix B.26, B.27 and B.28).

6.3.3

Following Environments

As outlined (in section 2.3.1.2.2) above, the diphthong /a/ has fewer restrictions on its

phonological distribution than the other two diphthongs under investigation. For

example, with respect to mono-morphemic words, it can occur before both voiced and

0

20

40

60

80

100

Labial (N = 1029)

Coronal (N = 2350)

Dorsal (N = 151)

Glottal (N = 5)

None – X (N = 85)

143

voiceless consonants and all stops, fricatives and nasals, except //, / / and // in word

final position30

. However, the only approximant and affricates that can follow /a/ are /l/

and //.

6.3.3.1

Voice

The first dimension for investigation is the phonological environment of following voice.

As previously mentioned, a particular linguistic phenomenon associated with variation of

(a) and (a) (see 4.3.3.1 for the analysis of the (a) variable) is that known as Canadian

Raising. This is a linguistic process in which the diphthongal nuclei of /au/ and /ai/ are

half-open, mid vowels when followed by a voiceless consonant in the same syllable, but

not in other contexts. Thus, pairs of words such as write [t]~ ride [raid] and knife

[] ~ knives [z] are derived (Wells 1982c:494).

The graph in 6.7a shows the overall percentages for the Mersea Island data and we can

see that there was a greater degree of raised and back []-type variants produced before

voiceless consonants, while contexts with following voiced segments and word-final

tokens produce more [a]-type variants. With respect to the Canadian Raising literature,

word-final variants should behave in a similar way to following voiced contexts and we

can see that this is the -case with the Mersea data.

30

However, it is possible in British English to truncate the name ‘Nigel’ to [na].

144

6.7

Overall percentage of (a) variants according to the voicing of the following segment

This pattern is also replicated with respect to the Museum speakers (6.8) albeit with the

raised backed unrounded []-types more common in the context of following voiceless

segments. In addition, both the Older (6.9) and Younger (6.10) data sets demonstrate

preferences in keeping with Canadian Raising type patterns and, even though it is not as

categorical as the rule suggests, it seems that these linguistic environments do have an

impact on the variant patterns found in the data.

0

20

40

60

80

100

(N = 59)

(N = 45)

(N = 657)

(N = 472)

(N = 2337)

(N = 24)

(N = 6)

(N = 20)


145

6.8 Percentage of (a) variants according to following voice by Museum speakers

6.9 Percentage of (a) variants according to following voice by Older speakers

6.10 Percentage of (a) variants according to following voice by Younger speakers

0

20

40

60

80

100

(N = 35)

(N = 27)

(N = 188)

(N = 23)

(N = 28)

(N = 0)

(N = 1)

(N = 0)


0

20

40

60

80

100

(N = 24)

(N = 18)

(N = 434)

(N = 328)

(N = 1184)

(N = 13)

(N = 1)

(N = 6)


0

20

40

60

80

100

(N = 0)

(N = 0)

(N = 35)

(N = 121)

(N = 1125)

(N = 11)

(N = 4)

(N = 14)


146

6.3.3.2

Manner

With the exception of following flaps, each environment clearly favours the standard-like

a-type variants. However, the percentages for this environment were only based on 9

tokens overall.

6.11

Overall percentage of (a) variants according to following manner of articulation

6.3.3.3

Following Place of Articulation

With respect to following place of articulation, whereas following vowels (in tokens such

as ‘triangle’ and ‘highest’) provoke a wider range of variants, it is only following dorsals

which promote the intermediate -type above the more raised and rounded -type

variants.

0

20

40

60

80

100

Stop (N = 1614)

Fricative (N = 695)

Nasal (N = 887)


Flap (N = 9)

Affricate (N = 1)

Vowel (N = 77)

None – X (N = 293)

147

6.12

Overall percentage of (a) variants according to following place of articulation

6.3.3.4

Following /l/ and the Status of ‘Island’

The following data compare the distribution of variants according to whether the

following /l/ is tautosyllabic with (a) (i.e. it is in the coda) or not (i.e. it is the onset of the

following syllable).

0

20

40

60

80

100

Labial (N = 770)

Coronal (N = 1619)

Dorsal (N = 79)

Glottal (N = 782)

Vowel (N = 77)

None - X (N = 293)

148

6.13

Overall percentage of (a) variants preceding /l/

The data show that variants with backed [] ~ [] nuclei rarely surface in this context,

while low back open [] ~ [] variants are preferred and, on the surface, there seems little

difference between the influence of /l/ when each phonological position is compared.

However, when the 107 tokens of ‘Island’ are separated from the ‘Non-Coda L’ data, a

quite different pattern emerges.

0

20

40

60

80

100

(N = 2)

(N = 1)

(N = 3)

(N = 41)

(N = 29)

(N = 142)

Coda L Non Coda L

149

6.14

A comparison of the overall percentages of (a) variants according to /l/ type

The data now demonstrate how, with the extraction of ‘Island’, Non-Coda L does not

exhibit any smoothed and backed variants. Instead, ‘Island’ appears to pattern with the

tokens which have a following /l/ in the coda (such as while and tile), particularly with

respect to the use of the backed monophthong []. This is demonstrated by the overall

comparison of diphthong variants versus monophthong variants:

6.15

Overall percentage of Diphthongal vs. Monophthongal variants for the token Island

The data suggest that a monophthong realisation of ‘Island’ is lexically conditioned, since

it does not conform to the expected [] monophthong before coda /l/. In addition, this

0

20

40

60

80

100

(N = 2)

(N = 1)

(N = 3)

(N = 41)

(N = 29)

(N = 142)

Coda L Non Coda L Island

0

20

40

60

80

100

Diphthong (N = 27)

Monophthong (N = 80)

Island

150

pattern could also suggest that the pronunciation of ‘Island’ is a type of iconic marker

within Mersea Island speech (this latter point will be discussed further in Chapter 9

below). Indeed, this pattern is reinforced in each age group. However, as 6.18 illustrates,

the Younger speakers seem to be adopting more diphthongal realisations for ‘Island’

compared to the other two age groups. This could suggest that the lexical effect of this

item on variation is beginning to be lost in Mersea Island English as it converges with the

expected pattern of (a) before a non-coda /l/. However, the limited token numbers for

the Younger speakers (9 tokens) would need to be expanded before this claim could be

established.

151

6.16

Percentage of (a) variants for Island by Museum speakers

6.17

Percentage of (a) variants for Island by Older speakers

6.18

Percentage of (a) variants for Island by Younger speakers

0

20

40

60

80

100

Diphthong (N = 2)

Monophthong (N = 7)

Island

0

20

40

60

80

100

Diphthong (N = 20)

Monophthong (N = 69)

Island

0

20

40

60

80

100

Diphthong (N = 5)

Monophthong (N = 4)

Island

152

6.3.4

LIKE

The analysis of the MIE data yielded 746 tokens of LIKE. These were coded according

to either their grammatical function (298 tokens) or their discursive function (448 tokens)

following D’Arcy (2007). Below are examples of these individual functions from the

Mersea Island data (note - the letters after each example represent the informant’s speaker

ID code):

Grammatical Functions

Lexical Verb: “we didn’t like it” (DS)

Adverb: “he looks like him” (AF)

Conjunction: “it sounded like they were singing” (LD)

Suffix: “factory-like” (DS)

Discourse Functions

Discourse Marker: “Like, my last recording was with this little girl” (SG)

Discourse Particle: “this water's like gonna electrocute everyone” (CL)

Approximate Adverb: “it sat there for ages like days” (TH)

Quotative: “all the children were like "are you recording?" (SG)

The following tables show the overall token count for each category and their

corresponding percentages.

153

6.19a

Overall number of (a) tokens for the token Like

Total

Grammatical Functions Lex. Verb 15 21 87 0 0 123

Adverb 30 19 122 1 0 172

Conjunc. 0 1 1 0 0 2

Suffix 0 0 1 0 0 1

Discursive Functions Disc. Marker 1 3 17 0 0 21

Disc. Particle 18 53 291 0 1 363

Approx. Adverb 0 1 1 0 0 2

Quotative 0 3 59 0 0 62

6.19b

Overall percentage of (a) variants for the token Like

Lex. Verb 12.20 17.07 70.73 0.00 0.00

Grammatical Functions Adverb 17.44 11.05 70.93 0.58 0.00

Conjunc. 0.00 50.00 50.00 0.00 0.00

Suffix 0.00 0.00 100 0.00 0.00

Disc. Marker 4.76 14.29 80.95 0.00 0.00

Discursive Functions Disc. Particle 4.96 14.60 80.17 0.00 0.28

Approx. Adverb 0.00 50.00 50.00 0.00 0.00

Quotative 0.00 4.84 95.16 0.00 0.00

The token numbers demonstrate that it is the lexical verbs, adverbs and discourse particles

which yield the greatest number of tokens throughout the data, while the percentages

show that the standard-like a-type variants are preferred for each functional linguistic

category. However, the data representation in (6.20) below demonstrate a little more

clearly how it is the grammatical functions of LIKE which, when considered together,

feature a higher percentage of more traditional -type variants than the more innovative

discursive functions of LIKE.

154

6.20

An overall comparison of (a) variants according to the functions of LIKE

With respect to individual age groups, the data from the Museum group only produced 9

grammatical tokens (8 of which were [] and one of which was []) and no discursive

tokens. However, when looking at the other two groups, there appears to be little

difference between the variant pattern of grammatical and discursive LIKE in the

Younger speakers (6.22). However, there is a slight difference in the data relating to the

Older group, particularly regarding the greater use of the intermediate -type variants in

discursive functions of LIKE (6.21).

0

20

40

60

80

100

Grammatical Functions (N = 298)

Discursive Functions (N = 448)

155

6.21

A comparison of (a) variants according to the functions of LIKE by Older speakers

6.22

A comparison of (a) variants according to the functions of LIKE by Younger speakers

6.4

Summary

This section has presented results from the analysis of () in the Mersea Island English

data. It has shown that there has been a change from traditional []-type diphthongs to

more open [a]-type diphthongs, and this transition is illustrated nicely by the spread and

use of variants (particularly the intermediate []-type diphthongs) by the Older speakers.

However, regarding the Older speakers, it seems that it is the Older males who are

0

20

40

60

80

100



0

20

40

60

80

100



156

retaining the traditional variants to a greater extent than females (and this contrasts with

the behaviour of the Older group regarding the () variable, see Chapter 8 below).

With respect to linguistic factors, we have seen that syllable type, preceding and

following manner of articulation and following place of articulation do not seem to have a

great impact on the type of variant produced. However, it has been shown that preceding

manner of articulation shares some similarities with the analysis of (a) (see sections

4.3.2 and 6.3.2 above) in that preceding dorsal consonants actively promoted lower [a]

nuclei and, thus, wider diphthongs. Once again, this could be one of the phonological

environments leading the change to [a]. Also, an analysis of following voice has shown

that a pattern similar to that of Canadian Raising is present, with more raised diphthongal

nuclei surfacing before voiceless consonants and more open, or lower variants surfacing

before all other environments.

The specific environment of // + /l/ was investigated and results show that, when coda

/l/ and non-coda /l/ data are compared, there appears to be no difference between them, in

terms of the variants used. However, when the token island is extracted from the non-

coda /l/ data, it has been shown that non-coda /l/ tokens favour a diphthongal variant

while coda /l/ tokens and island clearly favour monophthongal variants.

Finally, the investigation of LIKE has shown that, when LIKE functions grammatically, it

correlates with a higher percentage of traditional variants while discursive functions

correlate with more recent [a]-type variants. However, even though the Older group

shows a difference in this direction, there appears to be no contrast in the variant usage

for the Younger speakers.

157

Chapter 7


Variation of the Variable ()

158

7. THE HISTORICAL DERIVATION AND VARIATION OF THE VARIABLE ()

The following will be a discussion of the derivation of present day // in British English.

Included will be its historical sources, as well as data from historical and present day

dialectal studies, which will demonstrate the type of variation which can be found in

Essex and, more broadly, across the United Kingdom.

7.1

Historical Sources

The variable () represents the stressed diphthong // or what Wells (1982a) classifies as

the CHOICE vowel. Unlike /a/ and /a/, // did not develop as a result of the Great Vowel

Shift. Instead, present day // words are derived from Middle English /i/ and /ui/

diphthongs, and Wells notes that “all CHOICE words are believed to be ultimately loan

words, mainly from French” (1982a:209). Thus, those words which are derived from

Middle English /i/ include boy, oyster and noise, while those which are derived from

Middle English /ui/ include oil, boil (v) and join. In addition, Wells notes that in some

dialects those words derived from Middle English /ui/ developed into [i] or [i] which

resulted in these forms becoming homophones with those of the PRICE set (see below).

In addition to these sources, the CHOICE lexical set is completed by some lexical items

which contained Middle English /i:/ (and thus ‘should’ have developed into present day

/ai/ with the rest of the PRICE set) - for example, the final syllable of employ became

distinct from that of imply (Wells 1982a:209)

159

7.1.1

// and // - a Near or Temporary Merger?

Labov (1994) demonstrates, through literary sources and historical dialect observations,

that an apparent merger between the diphthongs // and // seems to have started around

the fifteenth century. However, even though there is literary evidence at this time, Labov

(1994:307) suggests that general recognition of this merger began to appear in the late

seventeenth century, when pairs such as loin ~ line and oil ~ isle were treated as

homophones.

Following the work by Nunberg, who traced the history and progress of change of the

modern // diphthong, Labov (1994) presents data suggesting that, in the Middle English

period, the // diphthong had two allophonic realisations: [ui], with a high back nucleus,

and [oi], with a mid back nucleus. As the nucleus of the [ui] allophone lowered and

centralised in the seventeenth century, it became close to the developing Middle English

as it progressed towards modern standard /a/. During this centralisation phase, the

allophones of // were affected differently according to phonological context. For

example, // diphthongs in pre-/nt/ environments (e.g. joint) were more likely to

centralise than those in pre-/l/ environments (e.g. boil). As a result, Labov notes that,

during this period, “the reflexes of ME and the originally high allophones of /oy/ were

[…] of close approximation”31

(1994:375). This can be illustrated as follows:

31

Note, Labov uses /oy/ to represent //

160

7.1 The development of /a/ and // demonstrating the potential overlap of phonemes

(from Labov 1994:376)

The point of overlap between // and //, illustrated by the diagram, is the point which

Labov describes as the diphthongs being “so close that they were judged as ‘the same’ by

most observers” (1994:376). However, he also notes that a number of words switched

diphthong classes during this time, indicating a type of merging before the phonemes ‘re-

separated’. By way of illustration, eyelet switched from // to /a/, while boil, joist and

groin switched from /a/ to //. This was also demonstrated in data from Suffolk by

Kokeritz (1932), where the items boil, noise, oil, spoil and poison were all variable

between [] and []~[] realisations.

By the end of the eighteenth century, the merger seemed to be in retreat but, by this time,

it had already attracted negative connotations and was associated with ‘common speech’.

161

However, the process of de-merging these two diphthongs continued and they remain

distinct in Modern English dialects. This, Labov (1994:308-309) notes, is true of a wide

range of British English dialects (including those representing the London vernacular,

Southern British dialects as well as Birmingham and Manchester dialects). The only

exception, Labov indicates, where the // and // merger “appears in current dialect

records [is] the county of Essex in the southeast” (1994:309).

Using data from the Survey of English Dialects, Labov extracted /a/ and // data from

seventeen locations across the county of Essex. Nine of these locations demonstrated a

merger between the diphthongs, while only two had distinct realisations for /a/ and //.

The remainder of locations, however, demonstrated separate realisations for /a/ and //,

but, phonetically, the differences between them were smaller, [] and [], respectively.

This was substantiated by Labov’s own analysis of data from the east Essex village of

Tillingham (which was fully merged in the SED data). However, spectrographic analyses

conducted by Labov suggested that the realisations of the ‘merged’ diphthongs were

phonetically distinct. This led Labov to hypothesise that, even though a merger was

perceived, “/ay/ and /oy/ were not completely merged in the eighteenth century, but only

closely approximated in a near merger” (1994:384).

162

7.2


The following section will focus on data from the selection of historical data sources as

outlined in 2.4 above.

7.2.1

Ellis (1889)

Even though Ellis does not give any specific commentary on this variable in Essex, his

data suggest that at least some parts of Essex retained the split in which () is

pronounced differently in tokens derived from Middle English /ui/. For example, the five

items which were present in the Essex word-list data predominantly show a rounded [o]

for the diphthongal nucleus in the words boy and voice, while an open [a] nucleus is

present in spoil, boil (v) and oil.

Lexical Item Transcription in Ellis

(1889)

Vowel translation of Ellis

using Eustace (1969)

Boy bo b

Voice wois ws

Spoil spa’il spl

Boil (v) ba’il bl

Oil a’il l

163

However, it is interesting to note that these forms do not seem to be homophonous with

the PRICE set in Ellis’ survey. Indeed, as noted above in section 5.3.1, the predominant

pronunciation of the PRICE vowel has a lower and backed rounded nucleus. As a result,

the PRICE set represents [], which contrasts with both sets above with respect to height

and lip rounding. Therefore, it seems that the above tokens (in which () precedes /l/)

are not homophones with (ai), but have remained distinct from the rest of the () set

through a lowering of the nucleus.

7.2.2


The dialect study conducted by Kurath and Lowman indicates that data from around

locations in East Essex and West Essex (towards London and Middlesex) include

derivatives of Middle English /oi/ and /ui/ surviving as distinct phonemes [o] and [],

respectively (Kurath and Lowman 1970:26). Thus, in these areas, point and oil rhyme

with pint and mile. However, this distinction is not represented in the data from the SED

(see below) suggesting that this dialect feature was receding at the time of these studies.

164

7.2.3

The Survey of English Dialects (SED) (Orton 1962)

The data from the SED, like that of Ellis (1889), is limited in the number of available

tokens. However, it seems that the spilt observed by Ellis is not present in the six

available tokens from SED’s data. For example, the following table shows the

pronunciations across three locations (note, only these words have been included as they

were represented by all three locations):

East Mersea Tiptree West Bergholt

Boiling bln bln bln

Poisonous pzns pzns pzns

Voice vs vs vs

The data from East Mersea, as well as that from Tiptree and West Bergholt, show

predominantly [] variants and occasional alternations with more centralised [] nuclei

and a raised, rounded [] variant (such as boys [b] in Tiptree and West Bergholt).

Thus, in East Mersea, both boil (v) and voice contain the present day standard [] form.

7.3

Modern Variation Studies

This section will examine the behaviour of () with respect to both British English

dialect studies and surveys as well as its behaviour in more isolated, insular varieties.

165

7.3.1

British English

Despite its intriguing historical development, there has been no systematic sociolinguistic

study of () within the British Isles. However, researchers have mentioned it in passing

when conducting broader dialect surveys.

For example, in Newcastle, Derby, Sheffield, Milton Keynes, Reading and Hull (see

Foulkes and Docherty (1999) for information on the individual studies), the CHOICE

vowel is noted simply as the standard [] variant. The wide geographical range of these

locations indicates the regional variation, with respect to this variable is minimal.

However, Trudgill (1999:130) notes that, in Norwich, the lexical item boil has a

noticeable PRICE vowel, though he remarks that this feature is very recessive. Other

variants of () in Norwich which Trudgill notes, range from [i], the most local form, to

the standard [].

Additional variation observed by Tollfree (1999:168) in South East London is that, in this

area, the CHOICE nucleus ranges between a rounded [o] and a more open [], while

younger speakers may use a very advanced or centralised nucleus, producing a diphthong

of an [o ] quality.

166

7.3.2

International Insular Variation

This section will specifically focus upon modern studies of insular communities. These

studies will highlight the behaviour of () in communities with unique socio-

demographic histories to see if any parallels may be drawn between them and the

behaviour of () in British English dialects.

7.3.2.1

The Falkland Islands32

Sudbury (2000) and Sudbury and Britain (2000), while including () in the overall

description of Falkland Island English (FIE), do not discuss the variable in detail.

However, the diphthongal quality [] reflects that of Standard British English and

Sudbury notes that this realisation “is consistent with all the historical sources and other

southern hemisphere varieties” (Sudbury 2000:180).

7.3.2.2

St Helena33

The quality of // in St Helenian English (StHE) was found by Schreier (2008; 2010a) to

have a close back and slightly raised [o ]-type nucleus (another indication, Schreier

claims, that StHE has taken part in some aspects of the Diphthong Shift) as well as a more

centralised [] offglide. In addition to this, Schreier (2008; 2010a) observes how one

speaker exhibited an // diphthong for certain // items, such as boil. This use of what

Schreier calls ‘an archaic dialect feature’ could be a remnant of the British inputs brought

to the island by the early settlers, possibly those who came from the southeast of England

32

See Chapters 3 and 5 for an overview of the Falkland’s historical development. 33

See Chapters 3 and 5 for an overview of St Helena’s historical development.

167

in the early part of the eighteenth century. However, Schreier suggests that we should

“probably not interpret too much here since this realisation is rare and unusual in StHE”

(2008:171).

7.3.2.3

Tristan da Cunha34

The // diphthong in Tristan da Cunhan English (TdCE) has a close back nucleus of an

[o]-type quality. However, unlike StHE, there is no raising of the nucleus towards [o ].

Therefore, Schreier suggests that this diphthong has not taken part in the Diphthong Shift,

which he claims has affected the realisation of /a/., see section 5.5.5 above.

The use of the // diphthong for // also features in TdCE, and Schreier (2010b)

observes that it is also reflected in some local place names. This suggests that this dialect

feature was transported to the island during its early colonisation and that the denseness

of the social structure has helped in its preservation.

7.4

Summary

The Standard English diphthong has been shown to have a range of historical sources.

Looking at historical data, it has been found that, in some parts of Essex, lexical items

which have derived // via Middle English /ui/ have adopted an [a]-type nucleus to the

diphthong, such as [spail] spoil. Even though this split has been attested in Essex and

other areas of East Anglia, it seems that this feature has receded in favour of diphthongs

with a more standard-like quality //. However, present day variation of this diphthong

34

See Chapters 3 and 5 for an overview of Tristan da Cunha’s historical development.

168

does not reflect a use of [a] nuclei. Instead, the variation studies presented here

demonstrate that the realisation of this diphthong may include a degree of nucleus

centralisation ([] ~ []) and raising ([o]).

The dialect evidence from British-colonised insular communities mentioned above

demonstrates that lingering mergers of some // items with the // diphthong is still

possible, but only in the most conservative speech on St Helena and Tristan da Cunha. In

contrast, FIE represents a British standard-like // diphthong, suggesting that any merged

forms transported to the island by early settlers have been lost.

169

Chapter 8


Linguistic Factors

170


The following sections will present the results generated for the variable () in Mersea

Island English. Initially, the social dimensions of age and gender will be explored before

attention is turned to an examination of linguistic factors.

8.1

() – The Variants

286 tokens of () were extracted from the data and were coded according to the

methodology outlined in Chapter 2 above. The four variants identified throughout the

data differ with respect to the degree of backness and lip-rounding:

~ ~ ~ 36

8.2

Social Factors

This section will outline the results pertaining to () with respect to the social factors of

age and gender.

8.2.1

Age

The data in 8.1 show the distribution of () variants according to the age of the speaker.

It can be seen that the standard variant [] is the majority variant for all three age groups.

35

Examples of () tokens according to phonological environment can be found in Appendix D (3a) and

(3b) 36

Note: even though this variant has been included in the subsequent analysis, it only surfaced 5 times. In

each, it was before a tautosyllabic /l/ in the items ‘oil’, ‘coil’ and ‘coils’.

171

However, it can also be seen that it is only the Museum and Older speakers who use the

most traditional rounded and centralised variant ([]).

8.1a


Young 0.00 4.76 95.24 0.00

Old 25.14 13.14 58.86 2.86

Museum 29.63 3.70 66.67 0.00

8.1b


While the older speakers retain almost the same level of [] compared to the Museum

speakers, they exhibit higher levels of the seemingly intermediate [] variant. This is

another possible indication of the transitional nature of this speaker group.

0

20

40

60

80

100

Museum (N = 27)

Old (N = 175)

Young (N = 84)

172

8.2.2

Age and Gender

When the data are broken down by both age and gender, it is the behaviour of the Older

females that stands out, since it is this group which uses the highest percentage of [] and

[] (33.7% and 18.6%, respectively) compared to any other group.

8.2

Percentage of () variants according to age and gender

Thus, it seems to be the Older females who are utilizing the [] form more than their

male counterparts. Indeed, this is a higher percentage than even the Museum speakers

(who were all male).

0

20

40

60

80

100

Male (N = 45)

Female (N = 39)

Male (N = 89)

Female (N = 86)

Male (N = 27)

Young Old Museum

173

8.3

Linguistic Factors

This section will discuss and present the details of () variation according to the

linguistic factors of syllable type, phonological environment and lexical item.

8.3.1

Syllable Type

When we look at the overall data, we can see that neither syllable type appears to

influence the choice of variant, except in the case of [] which only occurs in closed

syllables. However, as already noted above, these tokens were all in the context of coda

/l/.

8.3

Overall usage of () variants according to syllable type

Once the data are broken down to show the individual age groups, we can see that the

older groups show a distinct preference for [] in closed syllables compared to open

syllables (8.4).

0

20

40

60

80

100

Open (N = 144)

Closed (N = 142)

174

8.4

Percentage of () variants according to age and syllable type

This observation is also apparent among the individual age and gender groups. The male

Museum speakers (8.5a) demonstrate a preference for [] in closed syllables and this

preference is also reflected in the data for the Older females (8.5b). With respect to the

Museum data, even though there were only 5 tokens of () in closed syllables, it is

interesting to note that the 3 [] tokens represented the lexical item ‘boys’ (while the

remaining two were ‘join’). This observation and the seemingly special status of BOY

will be discussed in more detail throughout the remainder of this analysis.

0

20

40

60

80

100

Open (N = 36)

Closed (N = 48)

Open (N = 86)

Closed (N = 89)

Open (N = 22)

Closed (N = 5)

Young Old Museum

175

8.5a

Percentage of () variants used by Mersea males according to age and syllable type

8.5b

Percentage of () variants used by Mersea females according to age and syllable type

0

20

40

60

80

100

Open (N = 23)

Closed (N = 22)

Open (N = 46)

Closed (N = 43)

Open (N = 22)

Closed (N = 5)

Young Old

Museum

0

20

40

60

80

100

Open (N = 13)

Closed (N = 26)

Open (N = 40)

Closed (N = 46)

Young Old

176

8.3.2

Phonological Environments

When investigating preceding and following phonological environments, the correlation

between variant patterns and the lexical item ‘boy’ (together with its derivative ‘boys’)

which was previously alluded to with respect to syllable type, is reoccurring. For

example, due to the BOY effect, which sees a higher correlation of [] following the

labial stop in the items ‘boy’ and ‘boys’, () in word final position and before fricatives

show a higher percentage of []. Therefore, due to the biased nature of the following

phonological environment data, together with the distributional restrictions outlined in

section 2.3.1.2.3 above, only preceding phonological environments will be discussed here

to illustrate and establish the BOY effect. (For a break down of following phonological

environments plus graphs, see Appendix C.16-C.27).

8.3.2.1

Preceding Place of Articulation

The preceding place of articulation demonstrates that it is a preceding labial place which

promotes the rounded [] variant considerably more than any other category. This can

be seen in the overall data provided in 8.6 below. However, it is also interesting that the

3 tokens of preceding [sp] (all produced by older females) do not provoke an [] variant,

despite the labial quality of the adjacent [p]. This is perhaps an indication of the /s/-

cluster’s status as a complex segment, as opposed to two separate segments where the

second element has a full LABIAL specification.

177

8.6

The overall percentages for () variants by Mersea Island speakers according to


The correlation between preceding labials and [] is also apparent throughout each

individual age group, and it is the only environment in which young speakers were found

to deviate from the standard [] variant as a comparison of the graphs in 8.7, 8.8 and 8.9

will illustrate. However, it is not surprising, perhaps, that a preceding labial articulation

would promote a more rounded vowel articulation due to a coarticulatory effect which

spreads the LABIAL feature. However, the data for preceding manner also highlight a

particular correlation which cannot be explained so easily.

0

20

40

60

80

100

Labial (N = 157)

Coronal (N = 77)

Dorsal (N = 3)

Word-initial (N = 46)

[sp] (N = 3)

178

8.7 Percentage of () variants by Museum speakers according to preceding place of

articulation

8.8 Percentage of () variants by Older speakers according to preceding place of

articulation

8.9 Percentage of () variants by Younger speakers according to preceding place of

articulation

0

20

40

60

80

100

Labial (N = 10)

Coronal (N = 9)


0

20

40

60

80

100

Labial (N = 93)

Coronal (N = 43)

Dorsal (N = 2)


[sp] (N = 3)

0

20

40

60

80

100

Labial (N = 54)

Coronal (N = 25)

Dorsal (N = 1)


179

8.3.2.2

Preceding Manner of Articulation

With respect to preceding manner of articulation, it is preceding stops which consistently

promote [] and, once again, this is the only environment in which the younger speakers

seem to deviate from the standard. Figure 8.10 presents the graphic representation of the

overall figures for Mersea Island English (see Appendix C.10 for the complete data set

and Appendix C.28, C29 and C.30 for the individual age statistics).

8.10

Overall percentages of () variants by Mersea Island speakers according to preceding

manner of articulation

The data, together with data from preceding place of articulation, suggest that it is

preceding labial stops in particular which promote []. However, when we look at the

nature of the preceding stop data, it is found that, out of a total of 149 preceding stop

tokens, only 9 of these were not a preceding labial (6 were preceding /t/ and 3 were

preceding /k/), thus demonstrating once again the effect of a preceding labial on the data.

0

20

40

60

80

100

Stop (N = 149)

Fricative (N = 17)

Nasal (N = 20)

Affricate (N = 46)

Approximant (N = 5)

[sp] (N = 3)

None (N = 46)

180

8.3.3

The BOY Effect

Even though the analysis so far has shown preceding labials37

, in particular labial stops, to

be a favourable environment for the preservation of [], the analysis cannot stop there.

Consider the graphs in 8.11 to 8.13 below which illustrate the variant usage for the lexical

item ‘boy’ and any derivatives applicable to the individual age data sets. These graphs

illustrate that [] is the preferred realisation for both the Older and Museum groups. In

addition, it is the graph in 8.13 which shows that the only non-standard () tokens

produced by Younger speakers (given as 4.8% in 8.1a) occur in BOY tokens (3 ‘boy’ and

1 in ‘boys’). These data suggest that the non-standard variants surface across derived

forms of the BOY morpheme and not only the base form ‘boy’.

37

The data set for preceding labials consists of 157 tokens of which 139 were preceding stops and 94 were

BOY tokens.

181

8.11 Percentage of () variants across lexically derived BOY tokens by Museum speakers

8.12 Percentage of () variants across lexically derived BOY tokens by Older speakers

8.13 Percentage of () variants across lexically derived BOY tokens by Younger speakers

0

20

40

60

80

100

boy (N = 6)

boys (N = 3)

0

20

40

60

80

100

boy (N = 35)

boys (N = 33)

boy's (N = 1)

0

20

40

60

80

100

boy (N = 10)

boys (N = 3)

boyfriend (N = 1)

boyish (N = 1)

182

Therefore, the comparable status of BOY to all other () tokens can be seen in the

following. However, since the Younger speakers only have 4 non-standard tokens, they

have been omitted from the following.

8.14 A comparison of () variants between BOY tokens and all other tokens by Museum

speakers

8.15 A comparison of () variants between BOY tokens and all other tokens by Older

speakers

0

20

40

60

80

100

BOY (N = 9)

other (N = 18)

0

20

40

60

80

100

BOY (N = 69)

other (N = 106)

183

The asymmetry between BOY tokens and Other tokens is clear for both age groups. This

suggests that a type of lexical effect has attached to BOY and has thus enabled it to

preserve the traditional dialect feature to a greater extent than other words. Another

factor which may have caused the preservation of [] in these words is that boy is often

used in the performance phrase ‘my boy’ [ ] when Mersea Islanders are discussing

the speech of their elders or peers. For example, “oh, it’s always [m b] this or [m

bz] that” (from an Older Mersea speaker). Wolfram and Schilling-Estes (2006, for

example) note that a dialect performance is where speakers enhance the vernacular for

demonstration purposes and that performance phrases are specific rote phrases, which

highlight a number of dialect features. These practiced phrases act as preservers of

certain salient dialect features for native speakers and, thus, will help to preserve certain

forms within these specific contexts. However, even though the amount of data is

restricted with respect to token numbers, this effect seems to be receding as the Younger

group does not appear to have the [] variant to the same degree as the previous

generation, even in the context of BOY tokens.

8.4

Summary

The analysis of the () variable has shown that the standard diphthong [] is the

favoured variant across each age and gender group. However, once again the transitional

nature of the Older generation can be seen. This group not only has a relatively high level

of traditional [], but also the highest level of the intermediate [] variant and it is the

females in this group who are retaining the traditional forms more than the males.

184

Linguistic factors appear, at first, to play a role with respect to preceding place and

manner of articulation. However, even though the analysis initially shows that preceding

labials and stops act as the main preservers of the traditional [] form, a closer look

suggests that it is in fact the preceding labial stop associated with the lexical item BOY

which is the source of most [] forms within the data.

185

Chapter 9

Mersea Island English -

Sociolinguistic Considerations

186

9. MIE - SOCIOLINGUISTIC CONSIDERATIONS

This chapter will discuss the data presented in Chapters 4, 6 and 8 in more detail. The

variant patterns discovered in the MIE data will be considered alongside sociolinguistic

issues relating to linguistic change (such as change through contact), the creation of

iconic markers within a dialect, as well as language change being the product of more

‘natural’ linguistic tendencies and motivations. We will see how the use of certain

variants in dialect performance can act as preservers of traditional forms and may also

indicate whether certain phonological rules are encoded in the dialect grammar. In

addition, we will discuss how socio-demographic change can lead to the widening of

social networks and social practices and how this may be linked to the creation of supra-

local variants.

9.1

The Direction of Change

The data analysis of MIE has shown that, while the Museum speakers predominantly use

the older and more traditional forms of all three variables, the Younger speakers use more

of the innovative forms. However, even though there is variation in these age groups,

neither exhibits quite the extent of variation as the Older speaker group.

Though variation in a language or dialect can indicate a state of stable variation (that is,

variation which occurs between variants with none evolving to become the primary

variant over time), when it is indicative of change, age is the primary social factor with

which the variation correlates. Therefore, when looking at variation in sociolinguistic

data in either real or apparent time, change is indicated when a variant, which is least

187

common in the oldest age group, becomes more frequent as the age of the speakers

decreases. However, in order to ensure against age-grading patterns being interpreted as

change, apparent time data, which is supplemented by real time data, is more reliable.

The Mersea data incorporates both of these perspectives - apparent time data sets were

collected from the interviews of the Older and Younger groups while real time data sets

were added through the museum recordings and data from historical sources (such as

Ellis (1889) and the Survey of English Dialects).

The data from MIE also highlights the main path of change, with respect to the onsets of

the three diphthongs under investigation, from the oldest generation of informants (the

Museum speakers) to the youngest generation of informants (the Younger Speakers).

This can also be represented as follows:

9.1 The direction of change for the nuclei of Mersea Island //, // and // diphthongs

From this we can see the movement of the diphthong nuclei. Even though we might

claim (especially with regard to the behaviour of the back nuclei system) that there is a

type of chain shift involved, further analysis on the diphthong system as a whole would

need to be conducted to ascertain the exact nature of the shift. However, by looking at

the Mersea data, it seems that the Museum speakers were furthest advanced in the change

188

relating to the // diphthong in comparison to the // and // diphthongs (see Chapter 8)

and therefore, it might be proposed that it was the shift of the [] nucleus to [] which

prompted the lowering of // (from [] to []) as a merger avoidance strategy.

Regardless of what movement prompted these changes, the direction of change is not

what we would expect if the Mersea dialect was undergoing the Diphthong Shift (Wells

1982a) or the Southern Shift (Labov 1994). Both these chain shifts predict that, for

example, the nucleus of // would back towards [] while the nucleus of // would raise

towards []. However, the MIE data show no evidence of either process. Instead, the

reverse is true. A comparison of the oldest (Museum) speakers, as well as the data from

historical dialect surveys, with data from present day Islanders, suggests that the nucleus

of // has backed to become less centralised, while the nucleus of // has lowered from

the [] region. These contrasting developments can also be seen in the data from the

variable (a). Whereas, for example, the Diphthong Shift would represent a fronting and

raising of the // nucleus towards [] or [], the data comparison for Mersea

demonstrates that, in this dialect, the direction of change has been from [] to [].

With respect to the nature of this change, as previously mentioned, the data from all three

diphthongs showed a wider range of variation in the Older speakers’ group when

compared to the other two groups. This is not unexpected in a speech community which

is experiencing change in progress. During the change period, the initial stages are

relatively slow, with the innovative variants used infrequently (this could be said to

reflect the Museum speech). However, as the momentum of the change increases, it leads

to a robust period of variation where the change rapidly spreads through the speech

community before a cooling-off period, where the rate of progress slows once more, as

the change nears completion. It could be said that these two stages are reflected in the

189

nature of the Older and Younger speakers’ data, respectively. This type of momentum

change leads to a graphic pattern known as an S-Curve and can be demonstrated by the

following data from Chambers (2002:360)

9.2 An S-Curve: pattern of change for the variable (wh) in Central Canada

This graph illustrates the pattern of change for the variable (wh) in Central Canada. We

can see how the change from [hw] to [w] (in words such as which) begins very slowly in

the oldest age groups before undergoing a rapid rise in the use of the innovative variant

[w]. The tailing-off period begins in the two youngest age groups when the change is

nearly complete. Chambers (2002) explains that the S-Curve has been observed in a

range of linguistic diffusions and, with respect to (wh), even though the projection of a

total loss of [hw] by continuing the trajectory of the curve is scientifically logical, with

respect to linguistic behaviour, “linguistic variants that are well-entrenched in the

190

language, as this one clearly is, tend to linger” (2002:326). Therefore, a diffused change

may never reach its full conclusion. This may be because the diffusion is halted before it

can affect all linguistic environments or that, even though the older feature has lost its

social standing within a dialect community, it is still preserved as a feature of that

community through iconic phrases and relic forms (including, for example, street and

place names). Indeed, the preservation of forms in certain phrases and contexts can

prevent a change from reaching full completion and, as a result, the forms can acquire

particular types of social symbolism, both within and outside a community.

This point will now be discussed in relation to the patterns of Canadian Raising, BOY

rounding and ‘Island’ monophthongisation found in the Mersea Island data.

9.2

Performance and Enregisterment

Schilling-Estes (1998) defines performance speech as “that register associated with

speakers attempting to display for others a certain language or language variety, whether

their own or that of another speech community” (1998:53). In contrast to the type of

natural speech modern sociolinguistics strives to capture, it is the type of speech which is

above the level of consciousness for a speaker who, in a sense, chooses to manipulate

their speech in order to perform or emulate certain linguistic characteristics. This was

discovered in both the speech communities of Smith Island and Ocracoke (off the North

Carolina coast, USA) with respect to the variables () and (). Schilling-Estes and

Wolfram (1999:503) note that Smith Islanders manipulate and openly discuss the nucleic

quality of // diphthong but not the quality of the // nucleus, suggesting a greater

awareness of // variation at both speaker and community levels. Conversely, Ocracoke

191

speakers ‘performed’ variation of // but paid little attention to the variable qualities of

//. Indeed, Schilling-Estes’ (1998) analysis of a male Ocracoke Islander, Rex, shows

that during the recitation of a particular rote phrase (which includes the popular Island

phrase ‘It’s a high tide on the sound side’), he consistently manipulated and raised the

nucleus of the // diphthong towards the older, more traditional form but did not perform

that same action for the // diphthong, which was also contained within the domain of

the ‘performance phrase’ in the lexical item ‘sound’. Schilling-Estes (1998:63) goes on

to note that the ‘artistic performance’ accompanying Rex’s linguistic performance

encodes the persona of a “jovial old fisherman”. This suggests that there is a level of

social meaning attached to and implied by the performed linguistic structures and that

these social connotations are available for interpretation by those who are perceiving the

speech event as a whole.

This link between certain linguistic structures and encoded social values has been present

and discussed with respect to the notion of ‘dialect enregisterment’ (see for example,

Agha (2003), Beal (2009) and Johnstone et al (2006)). Agha defines enregisterment as

the processes through which ‘a linguistic repertoire becomes differentiable within a

language as a socially recognized register of forms” (2003:231). Therefore, by using a

specific linguistic feature, or set of features, a person can establish a link between those

features and, for example, a certain geographic location and/ or social group belonging to

the location. However, this kind of correlation can only be established if the speaker or

perceiver is aware (on some level) of the linguistic difference in the first place (that is,

members of a speech community cannot attribute distinct social correlations of linguistic

forms within their community if it is thought that everyone in it speaks the same way).

Therefore, variable linguistic features must be perceptually distinct or salient in order for

192

them to be initially recognised as ‘different’. Once linguistic features of a speaker or a

community have been tied to social aspects, such as locality or stylistic variation between

variants which are considered ‘local’ and ‘non-local’ (the latter most likely belonging to a

standard variety), they become available for dialect performances and overt discussion,

both within and outside the community.

The behaviour of // and // diphthongs in Ocracoke and Smith Island, mentioned

above, can illustrate how features may become enregistered within a community in

different ways. For example, since variation of the // diphthong is openly discussed as a

feature of Ocracoke English by native Islanders but not variation relating to //, this

suggests that it is the former which is enregistered for these speakers and not the latter.

As a result, // raising and backing is available for manipulation in performance phrases.

Conversely, the opposite can be said for Smith Island English, where // raising could be

said to be enregistered (and thus open for social comment between Islanders) and not //.

With respect to how the performance of enregistered features can reflect linguistic

structure, Schilling-Estes (1998:62) notes that the performance speech of her subject,

Rex, emulates the linguistic patterning of the variable () with respect to linguistic

context (that is, a raised // nucleus is more prolific in certain phonological environments

compared to others, see section 5.5.2 for more details). This indicates that aspects of

phonological conditioning of this variable must also be enregistered for this speaker.

This is also reflected in Chambers’ (1989) observations with respect to // and // in

Canadian Raising dialects. Citing and discussing Vance’s 1987 observation that

Americans, while prone to making overt comments regarding words with raised and

193

backed nuclei pronunciations of //, ignore words which are prone to // raising,

Chambers states that “/aw/-raising and /ay/-raising exist independently of one another in

the phonologies of regional accents” (1989:77). Therefore, if we accept that the raising

rules for these diphthongs are independent in our phonologies, this can help explain why

one may become enregistered and disassociated from the other.

However, these scenarios differ from the apparent situation of Mersea Island English with

respect to // and // variation. This will now be discussed below along with the

variation discovered in relation to the BOY and ‘Island’ lexical items.

9.2.1

Performance, Enregisterment and Mersea Island English

The data analysis presented in Chapters 4 and 6 demonstrates a correlation associated

with the phonological process of Canadian Raising. However, when both Islanders and

outsiders comment on their speech and use rote phrases38, they use raised // and //

nuclei consistently across phonological contexts and do not accommodate the Canadian

Raising rule during any performance or dialect comment. Therefore, it could be

suggested that it is only the raised nuclei of these diphthongs which are enregistered or

associated with this dialect and not any type of contextual phonological raising rule.

This could be further evidence to suggest that the phonological context correlations found

in the data are the result of natural trajectories and tendencies in the linguistic changes

from // to // and // to //, since the phonological rules have not become stabilised

or enregistered as part of the dialect. Britain (1991) discovered a strong Canadian

38

Note, this is based on both data from the interviews which make up the whole Mersea Corpus as well as

the researcher’s personal experience.

194

Raising pattern when analysing his data of Fenland speakers (although, this pattern was

only found for the variable () and not ()). Britain explains his data through increased

contact between dialect varieties: [] variants from the Western dialects, [] from the

Eastern dialects and whatever the native Fenland form was at the time. When these

varieties began to mix, due to increased contact opportunities, the Central Fenland variety

became more focussed as Western [] forms were reallocated to the context of pre-

voiced consonants while the Eastern [] forms were reallocated to pre-voiceless

consonants.

The motivation for this pattern of variant allocation can be found through the notion of

phonological naturalness. Laver (1994:445-447) presents a number of cross-linguistic

studies which demonstrate varying vowel durations in the context of following

consonants (that is, following consonants which are part of the same syllable rime) and

notes that, with respect to English, vowel durations before voiceless consonants are

shorter than those before voiced consonants. Working on this observation, Trudgill

(1986:155) proposes that, since a general tendency of most English varieties is the

production of longer vowels in pre-voiced consonant position, more open diphthongs

would then also be favoured in this environment (due to the greater amount of time

available for their articulation).

The MIE patterns for both () and () can therefore be attributed to natural trajectory of

change and variation according to phonological environment. If we accept that the

traditional MIE forms were // and // in all phonological positions, as contact

increased between Islanders and Mainlanders (through increased off-Island transportation

and a significant increase in residential Island housing, leading to large amounts of in-

195

migration), greater exposure to the more mainstream, innovative [] and [] variants

would have occurred. The subsequent change towards these wider diphthongs therefore

followed the more natural allocation path with the wider diphthongs becoming more

prominent in pre-voiced consonant positions and the narrower diphthongs more

prominent in pre-voiceless consonant environments. The Canadian Raising type pattern

uncovered in the data can therefore be viewed as the residue of changes from [] to [a]

and [] to [a], and is thus a linguistic constraint on change rather than a phonological

rule as it exists in Canada and the Fens.

The intertwining nature of phonological change and phonological structure was observed

by Weinreich et al (1968) who state that “linguists are naturally suspicious of any account

of change which fails to show the influence of the structural environment upon the feature

in question” (1968:172). Thus, part of the embedding problem outlined by Weinreich et

al suggests that a feature is ‘embedded’ in a linguistics matrix. Consequently, as one

features changes, other forms within the matrix are susceptible to change as well, even

though changes affecting other features within the matrix may take a number of years to

come about. However, in addition to the linguistic matrix of features which acts to unify

the linguistic system, or parts of the linguistic system, Weinreich et al caution that any

change is also embedded within a complex social matrix. Therefore, the interpretation of

language change data “depends upon the entire sociolinguistic structure” (1968:177), not

just observations with respect to the variant distributions across real and apparent time.

The linguistic conditioning on the variation and subsequent change regarding the

variables () and () is, therefore, similar to the conditioning factors of Neo-

Grammarian sound changes, where certain phonological environments may be affected

196

before others during change (see Labov (1994) for an overview and evaluation of types of

sound change). In this case, it was the more natural environments which led the change

and perhaps instigated the shift in the underlying phonemic structure, or matrix.

However, unlike the Fenland data discussed above, which show a strong and distinctive

Canadian Raising pattern for // due to the degree of focussing in the Central Fenland

areas, the Mersea data for both variables do not show a strongly focussed pattern. This

may be attributed to the relative demographic stability enjoyed by Fenland areas after the

initial contact event occurred, whereas the contact situation on Mersea has not been

followed by stability of this kind. As a result, the phonological processes governing the

raising patterns have not stabilised in the MIE speech community and the change is

working towards completion. Therefore, it seems logical to assume that this Canadian

Raising type pattern, which seems to be below the level of consciousness and social

comment, will not become a defining feature of this dialect and that the change affecting

these diphthongs will eventually affect all phonological environments.

However, even though there seems to be no phonological raising rule embedded in the

performance of // and //, according to following context, there seems to a possibility

of interacting phonological and lexical conditioning applying to the items BOY and

‘Island’ for // and //, respectively. These will now be discussed in turn.

With respect to the performance phrase ‘My Boy’, we can argue that the ‘my’ element is

rendered [m] due to an extended performance convention which affects any underlying

//. That is, all /a/ diphthongs are raised to [] during performance. However, it could

also be argued that a pronunciation of [m] is only triggered in the specific environment

of this phrase. Without evidence to strongly support either of these claims (due to ‘my’

197

not appearing in any other performance phrase context), the status of ‘boy’ in MIE must

also be discussed. Schilling-Estes (1998:61) notes that the prominence of the raised //

nucleus in pre-voiced consonant position is enhanced by its frequency in this context

within performance phrases, compared to the frequency of pre-voiceless consonant

positions. The same could be said for the context of // following a voiced bilabial /b/.

This particular labial segment is by far the most frequent throughout the data sets and

derived BOY forms were the most frequent lexical items with a preceding /b/. Thus, the

data show that, for all age groups, the preceding /b/ segment promotes more non-standard,

centralised variants in natural, non-performance speech and is the variant of choice in the

‘my boy’ performance phrase.

One possible reason for the preservation of this feature in this phonological context

(already advanced in section 8.3.3) is the spread of labiality from the /b/ to the following

vowel. However, we could also claim that, as a result of both this phrase’s prominence

and frequency of environment, these factors have helped enregister and associate the []

variant with traditional Mersea Island English and the context of preceding /b/ in

particular. Therefore, if it is the phonological rule which has been encoded (i.e. // is

fronted following /b/) we would expect to find the fronted centralised variant ‘performed’

after a /b/ even when it is not in the context of the ‘my boy’ phrase. Conversely, if the

phonological rule has not been encoded, the fronted centralised variant would not be

realised outside the context of the lexical item BOY. Unfortunately, the data collected

which represents the Mersea corpus does not hold the answer to this, as it does not

represent any speaker using elongated performances which might include the sequence /b/

+ // in another lexical context. However, since the fronting and centralisation of the

diphthong in BOY words is apparent in both natural speech and in the performance phrase,

198

it might be suggested that the retention of the non-standard traditional form is not simply

a by-product of its enregisterment, but rather it may be strengthened in both speech

contexts by a combination of natural phonological processes and the embedding of its

social representations.

In contrast to both // and // raising and // fronting, the monophthongisation of // in

the lexical item ‘Island’ does not appear to be an enregistered feature of MIE. The word

‘Island’ does not feature in any performance phrase, and monophthongisation in this word

does not feature in any comments of the dialect by the speakers who were interviewed.

The data presented in Chapter 6 demonstrate that, for the MIE speakers analysed in this

study, there is never a case where a monophthongal variant of an []~[] quality is

produced in the context of a following non-coda /l/ (for example, in words such as ‘pilot’

and ‘skylark’). Instead, these []~[] variants are only used where the following /l/ is in

the syllable coda and is therefore articulated as a dark /l/ (for example, in words such as

‘mile’ and ‘while’). However, within this set of monophthongal variants is the word

‘Island’. Even though this item represents // followed by a clear onset /l/, the data

clearly demonstrate that it patterns with that of the coda /l/ words and prompts

substantially more monophthongal variants.

Thus, these data suggest some kind of lexical effect on the variant conditioning in this

word. However, as mentioned above, this word does not seem to be a part of any

performance phrase known to the researcher (from both academic and personal

experience), and the monophthongal realisation in neither the word ‘Island’ nor in

ordinary coda /l/ words is commented on by Islanders in discussions relating to their

199

dialect features. These observations suggest that, unlike the features previously

mentioned in the section, this is not considered as either a linguistic or a social marker of

MIE. Indeed, the data presented in Chapter 6 show that the Younger speakers are losing

the monophthong as a variant in ‘Island’ and are instead converging it with the pattern of

the other non-coda /l/ words. It must be noted that there were only nine tokens of ‘Island’

in the Younger speakers’ data and so this assertion is only tentative based on such limited

data.

With this in mind, it may be possible to argue that either the monophthong was extended

to the non-coda /l/ ‘Island’ from the smoothing environment of the coda /l/ (perhaps as a

result of the frequency of this word in the particular society), or that there has been a

change from a monophthong to a diphthong in the context of following /l/, with the clear

/l/ contexts progressing through the change ahead of the monophthong preserving dark /l/

contexts. If this latter scenario was the case, the relic form [] could have been

prevented from joining the change due to its frequency of use in the speech of Islanders.

This scenario, if correct, could then be claimed to be spreading through the Younger

generation speakers. Thus, even though this generation still exhibits non-standard

behaviour in the form of 44% of monophthongs for ‘Island’ (compared to the 77% of the

Older speakers) and 54% monophthongs before regular coda /l/ words (compared to 74%

in the Older speakers’ data), it still represents a decline in []~[] variants.

However, previous linguistic research which has considered the effect of frequency on

change (such as Bybee (2002; 2006)) has suggested that, as a general rule, high frequency

words will change at a faster rate, compared to those of lower frequency with respect to

changes linked to phonetic processes. Conversely, changes associated with grammatical

200

or analogical change show high frequency items as inhibitors of change and thus

generally act as preservers of traditional forms. Therefore, since the case of variation

between a monophthong and a diphthong would be considered as a phonetic change, we

would expect a frequent word (with respect to the social and geographic context of

Mersea Island speech) like ‘Island’ to be at the forefront of change, not acting as the

preserver of the, possibly, older form. As a result, it may be a case for arguing in favour

of the independence of place names in frequency calculations.

Perhaps this apparent ‘anti-frequency’ effect could be attributed to a type of social

saliency which has attached to the [] form in the past, with it becoming so

entrenched as to enregister the feature as part of the local speech pattern. Thus, a

weakening of the enregisterment would mean that current generations do not have this

level of social attachment to the feature and, as a result, it is gradually being lost in MIE

speech. Therefore, we could propose that, while enregistered and salient features of a

dialect are subject to inter-speaker and inter-community variation, they are also subject to

inter-generation variation as well.

9.3

Contact, Diffusion and Supra-localisation

It has already been noted earlier in this section how the demographic structure of the

Island has changed so that Islanders now have greater opportunities to be in contact with

non-Islanders. Thus, where there once existed a strong natural barrier to inter-regional

communication for the Island, this barrier has been overcome and its effects have been

weakened, so that contact between mainland varieties and the Island variety has been

strengthened. This strengthening of contact between Islanders and non-Islanders has led

201

to a weakening of community network structures for the Island inhabitants. Even though

there is a great diversity in the network structure of Island residents (primarily depending

on whether they went to the local primary school and how involved a resident chooses to

be in Island life and traditional Island activities), the compulsory act of schooling young

Island children on the mainland from the point of secondary school onwards (which was

not the case for the Older generation speakers in the study) means that from the age of 11

years, all Island children immediately have their social networks opened up, loosened and

geographically expanded.

Discussing the process of dialect levelling and supralocalisation, Britain (2010) defines

the notion of a ‘region’ with respect to the classification of ‘space’. Basing much of his

discussion on Allen et al (1998), Britain notes that a region cannot solely be defined by

physical geographical space or politically defined borders. Instead, they are areas which

are formed by social practice. Thus, the consistent routinised movement of individuals

leads, over time, to the creation of spatialised patterns, and it is the domain of these

patterns which leads to the emergence of ‘places’ and ‘regions’. The perception of

geographical space as something which is less physically defined and more socially

defined automatically puts emphasis on the role of social practice, particularly routine

social practice (such as travel to work and school), and enables us to explore processes

immediately above the level of the ‘local’. Consequently, the term supralocal may be

used to denote processes that are above the regional level of analysis – for example,

changes taking place across a county relative to, say, a large town within that county

(Britain 2010).

202

Studying the diffusion of innovative linguistic forms at this more micro-level of society

can be done using observations of social network structure and strength, with respect to

both individual informants and general observations of a speech community as a whole.

Social Network theory, as proposed by Milroy (see Milroy (1987), for example), suggests

that the closer and denser your network ties, the less likely you are to be influenced by

externally derived linguistic innovations. Therefore, people living in small, tight-knit

communities, whose social and professional activities are confined to a limited area and

limited individuals, are less likely to be exposed to and accommodate with innovative

forms in their linguistic repertoire and more likely to reinforce traditional and perhaps

more marked forms. Conversely, the more open your social network, the more exposed

you become to linguistic innovations and the more likely you are to accommodate to a

variety of new linguistic features. As a result, those with open, loose networks are more

likely to accept innovative forms into their personal linguistic repertoire.

Thus, the gradual ‘opening-up’ of the Mersea Islanders’ social networks has meant the

speakers within the community have become more susceptible to the comparatively

innovative forms used in mainland varieties.

Indicators of the geographical widening of network structure can be seen in a number of

ways on Mersea Island. When the Older informants were growing up, education was

restricted to the Island school and employment was mainly provided by local Island trade.

In addition, due to the limited availability of transport on and off the Island, travel to the

mainland was typically restricted to those who had family living there. Conversely, the

Younger generation speakers automatically receive their secondary and tertiary education

off the Island and are mainly employed in mainland jobs. The immediate effect of this is

203

Popluation of Mersea Island based on census data from 1801 to 2001

0

1000

2000

3000

4000

5000

6000

7000

18

01

18

11

18

21

18

31

18

41

18

51

18

61

18

71

18

81

18

91

19

01

19

11

19

21

19

31

19

51

19

61

19

71

19

81

19

91

20

01

Date

West Mersea

East Mersea

a geographical broadening of network ties and this will inevitably lead to structural

weakening of personal networks. It is also a possibility for an Islander to construct and

establish, as early as their teens, dual networks – an Island network and a mainland

network, in which some or no members will overlap. For example, if a network group of

children are schooled together on the Island and then go to different (or even the same)

secondary school, the original Island ties may still be preserved by travel on the school

bus, but they will also have ties with mainlanders while at school and may socialise with

them outside school as well. These types of dual structures may well be preserved and

created in later life, as well when Islanders enter employment off the Island. It is these

types of bridges between groups which, Milroy (1992) proposes, can act as bridges for

incoming innovations to be transmitted from one group to another.

Another influence on the Island’s network structure is likely to be the substantial rise in

population over recent years. The following graph (already shown in Chapter 1 above) is

repeated below to demonstrate this.

9.3

The population of Mersea Island, based on Census data from 1801 to 2001

204

The sheer rise in population creates greater opportunities for interaction with those

incoming residents, who are likely to represent a wide range of dialect backgrounds.

However, it should be noted that increased population would not automatically signal the

weakening of network ties in a community. Individuals tied to Island life may still

restrict their interactions with those others who are deeply connected to the community

through family history, employment and social activities. It would be these networks that

would remain dense and multiplex and would continue to reinforce traditional linguistic

forms.

Another point to consider with respect to the regional spread of network ties is how

mobile Islanders may be in their day-to-day lives. The community itself has many shops

and amenities to provide for the population (such as garage mechanics, a doctor’s

surgery, a dentist and a vet). Therefore, the graph below illustrates the approximate

distance that a member of the Mersea workforce travels to work, compared to the same

data from Colchester residents and the overall statistics from the East of England

(previously presented in section 1.4 above).

205

9.4

a) Percentage of the West Mersea workforce population (aged 16-74) according to

distance travelled to work, from the 2001 Census


Works mainly at or from home 11.85 9.16 9.44

Less than 2km 17.85 20.75 20.06

2km to less than 5km 2.38 22.84 16.96

5km to less than 10km 3.83 12.33 13.73

10km to less than 20km 39.56 9.05 14.73

20km to less than 30km 3.22 6.38 7.80

30km to less than 40km 4.36 3.33 4.17

40km to less than 60km 1.85 1.94 4.22

60km and over 7.89 9.01 3.53

No fixed place of work 6.68 4.91 5.07

Working outside the UK 0.44 0.26 0.23

Working at offshore installation 0.10 0.05 0.06

b)

Percentage of the West Mersea workforce population (aged 16-74) according to distance

travelled to work, from the 2001 Census

0

10

20

30

40

50

Works

mainly at or

from home

Less than

2km

2km to less

than 5km

5km to less

than 10km

10km to

less than

20km

20km to

less than

30km

30km to

less than

40km

40km to

less than

60km

60km and

over

No fixed

place of

work

Working

outside the

UK

Working at

offshore

installation


The data demonstrate how just over 64% of the Mersea working population travel off the

Island, while the highest percentage of the Mersea workforce travel between 10 - 20km

(6.2 – 12.4 miles) to work (which is likely to represent travel to Colchester and its

206

surrounding areas). Conversely, only 29.7% of Islanders work from home or on the

Island (the latter group being represented by the less than 2km (1.2 miles) category).

We could suggest, therefore, that a typical Islander is likely to centre not only their

educational networks on the mainland (at least from secondary education), but also their

professional networks. As a result, it seems logical to suggest that a modern Islander’s

typical social network will exhibit ties not only on the Island, but in the surrounding areas

on the mainland, leaving the once isolated Island dialect more susceptible to influences

from these surrounding regions.

Britain (2002c), with respect to movement within geographical locations, states that

“Intra-regional mobility, whilst breaking down networks and routines at the very local

level, reinforces supra-local structure” (2002c:62). Britain refers to these areas which

have undergone regional levelling as ‘regional koines’ as they represent levelled, supra-

local varieties which act as larger areas of less geographically internal linguistic diversity,

since they represent an amalgam of locales, which were once a lot more linguistically

diverse and distinct. This process of dialect supralocalisation is also referred to as

‘regional dialect levelling’ which Torgersen and Kerswill define as “the reduction of the

number of variants following speaker accommodation through face-to-face interaction”

(2004:26).

The face-to-face interaction between speakers of dialect regions is also central to the

notion of macro-level linguistic diffusion although, in contrast to levelling which may be

considered on a micro-level, the geographical area in which diffused forms spread can be

much greater. Linguistic diffusion of innovative forms (see, for example, Britain (2002a)

207

and Trudgill (1986)) involves the introduction of a novel linguistic form from another

community in which that feature is the norm. The driving force behind diffusion types is

necessarily contact between speakers from each community (that is, speakers with a

linguistic form coming into contact with speakers without that same form, or vice versa).

Indeed, Britain (forthcoming) states that diffusion models supposedly reflect the

geographies of interaction and represent who speaks to whom and where. However,

Britain also outlines general weaknesses of contemporary diffusion models as relating to

their failure to adopt a richly socialised and interactional perspective relating to the spaces

across which features diffuse.

As linguistic forms diffuse, they may supplant existing forms or create intermediate,

compromised forms (see Britain (2002c) for more details). In addition, newly diffused

innovations may be re-evaluated with respect to their social connotations, and this could

have a direct influence upon the variation and interactional usage between local features,

newly diffused features and standard variants (see, for example, Docherty et al (1997) and

Milroy et al (1994) for discussions of this relating to the variable (t) in Newcastle).

Taking these points into consideration, it stands to reason, therefore, that linguistic

features which are most compromised during contact situations are those which are

distinct between the varieties which come into contact, since they will be perceptually

more distinct. Thus, if we propose that the features which are marked, both perceptually

and socially, are those which are the most at risk from levelling processes when contact

situations arise, the diphthongs under investigation are prime candidates for change, since

they are susceptible to social comment, evaluation and are treated as markers of the MIE

dialect.

208

However, two possible paths of explanation are that:

1) since the new incoming variants are those associated with the standard variety, the

change is influenced by this standard ideal and is therefore assimilating to it.

or

2) due to the contact situation (which has been developing and increasing between

Islanders and non-Islanders over the last 50 years or so), the linguistic change is

influenced not only by the incoming innovative variants, but also by the natural

motivations of the underlying phonological system. Therefore, the motivations

which cause a movement towards less-marked forms are perhaps triggered by the

weakening of strong, linguistically reinforcing, network ties.

Indeed, the former suggestion cannot tell the entire story, as we would be hard pressed

and perhaps irrational to devise a theory as to why emulating the standard form became

so important to these Islanders at a time when their local identity was so under threat.

However, if we suggest that supra-local forms of these diphthongs were adopted as a

result of contact and the creation of a supra-local dialect region, we might then be able to

theorise that the reason behind the generation and adoption of the innovative forms was a

preference for a more harmonic and ‘natural’ diphthong system in the underlying

phonology. Thus, a combination of approaches might be advanced to suggest that

phonological motivations allowed the innovative [], [] and [] forms to be generated

and awareness and exposure to the standard (through contact and off-the-Island

schooling) allowed these variants to become established within the dialect.

It is this point of phonological naturalness which will be discussed at greater length in the

next chapter where an integrated model of phonological derivation and sociolinguistic

conditioning factors upon variation will be presented.

209

Chapter 10

Socio-Phonology: the construction

of an integrated model

210

10. Socio-Phonology: the construction of an integrated model

This Chapter will explore the results from the Mersea Island data analysis (presented in

previous Chapters) in relation to phonological theory. The application of phonological

theory to this sociolinguistic data will allow the phenomenon of dialect change to be

explored from a different perspective. For example, the data presented in the Chapters

above highlight a shift from [] to [], [] to [] and [] to [] between the Museum

and Younger speakers. However, even though these changes (as well as specific patterns

of variation) have been discussed with respect to change, contact and performance, these

have not addressed the issue of what is happening in the underlying structure of the

dialect and the changes to the grammars of the dialect’s speakers. Furthermore, an

exploration of phonological theory can help inform an investigation of variation and

change by highlighting any linguistic constraints of influences upon the variation as well

as any factors which are likely to shape the trajectory of change.

Both Optimality Theory and Dispersion Theory will be introduced, and it will be argued

that they are effective tools for dealing with issues relating to the direction of change we

have seen in the Mersea Island diphthongs. Using a combination of these theories,

together with sociolinguistic constraints on the data, an integrated socio-phonological

model will be constructed and presented as a way of illustrating the interface between

Sociolinguistics and Phonology.

211

10.1

What is Optimality Theory?

Optimality Theory, as originally proposed by Prince and Smolensky (1993; 2004), is an

output-based model of language39

. Although it is generally associated with phonology, it

has found applications in other domains of language and, compared to previous input-

driven models40

, such as that presented in Chomsky and Halle’s (1968) Sound Pattern of

English, Optimality Theory allows for a graphical evaluation of not only the resulting

surface form, but also the other possible contenders which may be logically derived from

the underlying form. By way of illustration, 10.1 demonstrates the conventional lay-out

of an OT tableau.

10.1

/Input/ Constraint 1 Constraint 2 Constraint 3

Candidate 1

Candidate 2

In OT, the underlying representation of a linguistic unit is represented as the input.

Therefore, the input needs a mechanism which will transform it into differing possible

realisations. This list, or set of possible outputs (which includes one completely faithful

to the input) is produced by what is known as the Universal Generator, abbreviated to

GEN. McCarthy states that the phonological GEN “performs various operations on the

input, deleting segments, epenthesising them, and changing their feature values”

(2008:16). Therefore, GEN defines the range of the candidates which will compete with

each other through a particular constraint ranking. However, the inherently unrestricted

nature of GEN means that each set of candidates which can be derived from a single input

39

Thus, the input and its sequential derivation towards an output, through rewrite rules, is not the primary

focus. Instead, a number of possible outputs (including one which is identical to the input) are generated

and their performance is compared in relation to each other, not in relation to how they compare to the

original input. 40

Therefore, the focus is on the input and the application of transformation rules in order to generate an

output.

212

is infinite. McCarthy, acknowledging this problem, states that “the grammar as a whole

does not overgenerate because the constraints filter the contents of the candidate sets”

(2008:11). Therefore, by limiting or filtering the outputs of GEN, we do not end up with

candidate sets which include what could be considered as illogical or unreasonable, with

respect to the language being modelled (such as a competing candidate [] for an input

//).

However, the nature of the input itself is not without controversy. Prince and Smolensky

(1993:209; 2004:225) originally propose that, through the notion of ‘Richness of the

Base’, languages do not vary with respect to the structure of their underlying lexicons.

Therefore, the ideals behind richness of the base are that languages do not differ with

respect to their inputs (that is, the structures of their inputs are universal) and that the

diversity of forms in the base (or input) are such because there are no language-specific

restrictions in operation at this level.

In the tableau above, the input, which represents the underlying form, is provided in the

top left cell and possible outputs which may be derived from it are listed in the column

below. Along the row to the right of the input are noted the various constraints imposed

by the language which are applicable to a particular analysis. Constraints provide certain

criteria or conditions which the surface form must meet in order to be selected, and are

ranked in order of importance – the further to the left a constraint is placed, the more

important it is that a candidate does not violate its conditions. The solid line between

Constraints 1 and 2 reveals that there is a ‘dominance relation’ between Constraint 1 and

Constraints 2 and 3, meaning that Constraint 1 is the most significant in shaping the

phonology of the language and cannot be dominated. In addition, while we can say that

213

Constraint 1 outranks Constraints 2 and 3, the dotted line between Constraints 2 and 3

means that they are of equal status when it comes to the evaluation of possible output

forms. An alternative way of showing this relationship would be as follows:

10.2

Constraint 1 >> Constraint 2, Constraint 3

The double ‘>>’ shows that Constraint 1 is more dominant than the other two, while the

use of a comma to separate Constraints 2 and 3 shows that they are unranked in relation

to one another.

It is these constraints which act as the evaluators to ultimately allow the grammar to

select the most optimal surface form. This is done by assessing the number of violations

each potential candidate incurs. For example, if a constraint stipulates that no voiced

consonants may surface, then a violation mark will be allocated for every voiced

consonant in a potential candidate. To demonstrate this, if, for example, a language had a

markedness constraint (see below for details) such as *VOICED CONSONANTS which was

imposed on its inventory, a violation mark is added to the appropriate cell each time a

voiced consonant featured in a potential candidate. The following displays what

violations would occur for the input /bed/ and some of its possible outputs:

10.3

/bed/ *Voiced Consonants

bed * *

bet *

ped *

pet

214

Thus, in this example, [pet] would be the winning candidate from this candidate set as

neither consonant is voiced. The winner is then, typically, indicated by ‘’.

Therefore, one advantage of this model is that it allows for flexibility within our personal

grammars, particularly with respect to constraints which are unranked in relation to each

other (as illustrated in 10.1 above). In OT, constraints are intended to be regarded as

universal, and they can be rearranged language-specifically in order to accommodate

differing surface forms, which may derive from the same underlying input. This contrasts

with rule-based systems which cannot accommodate such variation in this way. For

example, a rule-based model may represent the following:

10.4

A → B / V ____ # 41

Thus, the input is fed into the rule and A would be transformed when positioned in the

linguistic context of VA# so that the surface form would be VB#. In order to change the

surface form of the underlying input, say in the context of language change, a new rule

must be written or rule ordering must be changed. However, if we were to represent this

in an OT framework, it could be represented as the following:

10.5

/VA#/ Constraint 1 Constraint 2 Constraint 3

a. VA#

b. VB#

c. V #

41

Note: V = Vowel and # = a word boundary

215

In this example, it is the nature and order of the constraints which determine the surface

form. The first candidate (a) is recognised as the faithful candidate, as it is a replica of

the input. The second candidate (b) allows for the transformation A→B, as given in 10.4,

while (c) has deleted the segment between the vowel and word boundary. (Note that fatal

violations, which cancel out a candidate’s chances of surfacing, are marked with an

exclamation mark ‘!’ following typical OT conventions)

10.6


a. VA# * !

b. VB# *

c. V # * !

In this tableau, VB# (b) is the winning candidate and thus the surface form because it

only violates the lowest ranked constraint. However, if we were to reorder the constraint

ranking so that Constraint 3 now outranks Constraint 2, the winning candidate is now V #

(c).

216

10.7


a. VA# * !

b. VB# * !

c. V # *

Therefore, it is possible to demonstrate variation in the same graphical representation. An

advantage of this is that the differences between languages, and varieties within

languages can be demonstrated by using the same basic components (constraints) in order

to explain their differences, as opposed to the construction of individual rules or

independent rule-ordering sets which accommodate contrasts. Indeed, Prince and

Smolensky (2004:4) note that it is Universal Grammar which provides individual

languages with the same set of general constraints and it is the way languages rank these

constraints which establishes the primary differences between them. Conversely, rule-

based systems rely on, for example, rule creation and simplifications, as well as feeding/

bleeding and counter feeding/ counter bleeding relations42

in order to create language and

dialect diversity and change (see, for example, Kiparsky (1968) for a discussion of these

processes).

However, the application of optimality theory to certain types of phonological processes

and developments has not been without its critics. McMahon (2007) observes that cases

which involve degrees of opacity are cases which provide the greatest difficulties for the

mechanisms of OT. In this context, the term opacity refers to a phonological rule which

42

If a rule, A, generates forms which will subsequently be subject to another rule, B, then we can say rule A

feeds rule B. However, if rule A were to take away or restrict rule B’s potential inputs, then we can say that

rule A bleeds rule B. When the rule orders are reversed, this is counter feeding or counter bleeding (Clark

et al 2007:148)

217

prompts a surface form which is not expected in a given phonological context. This is

formalised by Kiparsky (1973, 1976 - cited in McCarthy (2007:11) as follows:

Opacity

A phonological rule P of the form A → B / C___ D is opaque if there are surface

structures with any of the following characteristics:

a. instances of A in the environment C___ D,

b. instances of B derived by P that occur in environments other than C___ D,

or c. instances of B not derived by P that occur in the environment C___ D.

Therefore, in order to get any of the outputs in a, b or c, more than one derivational step

must be taken. In a Classic OT framework, a constraint ranking which prompts the

output CBD would need to be constructed to reflect the original rule. However,

additional language specific constraints (such as those stipulating specific lexical items)

would be needed if the output was, for example, CAD instead. Indeed, McMahon notes

that “the more opaque a phonological process is, the harder it is to model in OT without

considerable extension of the permitted constraint types and theoretical machinery”

(2007:343). Thus, in order to derive an OT framework which reflects opaque processes

(such as vowel alternations which result from the interaction of phonology and

morphology43

), novel constraints (which may or may not be proposed as part of the

universal constraint grammar) are introduced, or adapted frameworks are constructed and

applied to problematic data, such as the recent theory of Candidate Chains (as proposed

by McCarthy (2006; 2007)). On the other hand, McMahon (2007:357) does observe that

classic OT’s failure to represent opacity does allow for the future development of the

theory (although she stipulates that its development will depend on how people approach

problematic data and the strength of their argumentation).

43

An example of this would be the vowel alternations involved with, what McMahon (2007) refers to as the

Modern English Vowel Shift Rule (which can be seen in words pairs such as divine ~ divinity and sane ~

sanity)

218

10.1.1

The Nature of Constraints

A central feature of Optimality Theory is that every constraint has the potential to be

violated. Therefore, it is a matter of dominance relationships, as opposed to the rigidity

of traditional rule-based models where the interaction of rules needs to be regulated so

that their order and interaction yields only the desired output. These constraints are

described as belonging to two broad categories – Faithfulness constraints and Markedness

constraints.

10.1.1.1

Faithfulness Constraints

Faithfulness constraints, in their most basic sense, are those which prohibit any

differences between the input and the output. Therefore, the input specifications must

remain intact and unaltered in the surface form. The implication of such constraints helps

to force contrasts in the surface forms. Indeed, McCarthy (2008) observes that “no other

theory of language has anything like OT’s faithfulness constraints” (2008:13). In

addition, he adds that the actual notion of faithfulness constraints could only make sense

in a theory such as OT which allows for constraint violations. Two examples of active

faithfulness constraints are those known as DEP and MAX and these guard against

epenthesis and deletion, respectively. The constraint DEP, which is short for

DEPENDENCY, requires every segment in the surface form to be represented in the input

and, therefore, the output ‘depends’ upon the input as the source of its components.

When DEP is active in a constraint ranking, one violation mark is usually assigned for

219

every segment in a potential candidate which has no direct correspondent in the input

(that is, each segment which has be generated through epenthesis).

Conversely, the constraint MAX, which is short for MAXIMALITY, acts as an anti-deletion

faithfulness constraint. It stipulates that segments in the input must be maximally

expressed in the output. Thus, any segment in an input which does not have a direct

correspondent in the output (due to its being deleted) prompts a violation mark to be

allocated to that output candidate.

In order to see these two types of constraints in operation, consider the following

examples of nasal stop clusters in English:

In English, word-final nasal-stop clusters must share a place of articulation (such as in

bent [bnt], sank [sk] and pump [pmp]). If we have an input /snk/ in which the

alveolar nasal /n/ is paired with the velar /k/, the following candidates may be advanced44

:

a) Regressive assimilation resulting in a change from /n/ to []

b) Insertion of a vocalic segment between /n/ and /k/

c) Deletion of the word-final /k/

d) Deletion of the /n/

In order to generate the correct output of [sk], the tableau in 10.8 could be constructed

using the following constraints:

Faithfulness Input-Output (FAITH I-O): All specifications of the input are

preserved in the output

Maximality (MAX): All segments in the input must be represented in the output.

No deletion.

44

Note: this is only a sub-set of possible outputs for illustration purposes

220

Dependency (DEP): All segments in the output must be represented in the input.

No insertion.

Agree PLACE: The elements of a consonant cluster must share a place of

articulation.

10.8

/snk/ DEP MAX AGREE PLACE FAITH I-O

snk *!

sk *

snk *! *

sn *!

sk *!

Tableau such as these are there created as a type of post hoc rationalisation as it is only

the surface form which is available for analysis. Thus, once a constraint ranking is

established for one output, it can then be tested against others and used to predict how the

grammar would process particular inputs.

Thus, we can see from this example how Dep and Max guard against the construction of a

new syllable ([snk]) and the reduction of the nasal cluster ([sn] and [sk]),

respectively. The faithful candidate ([snk]) violates Agree PLACE which, since its

ranking dominates the lowest ranked Faith I-O constraint, the input cannot surface in

favour of the assimilated candidate [sk]. Once this ranking is in place, it can be tested

against other items with nasal clusters to check its validity and be used to predict the

surface forms of proposed nasal cluster inputs.

However, it must be noted that this example uses the constraint FAITH I-O and not an

IDENTITY constraint, which is an alternative segmental faithfulness constraint. In this

example, the constraint FAITH I-O is generalising over the collective features represented

by the segments in the input and ensuring that these features are not altered (such as the

221

alveolar nasal’s transformation to the velar nasal). However, the family of IDENTITY

constraints (IDENT for short) allows for a constraint representative of each distinctive

feature to be included within the constraint hierarchy. Thus, we can define it as follows:

INDENT (F) – Correspondent segments are identical in the feature (F)

(See McCarthy and Prince (1995:16))

Therefore, since the place feature is being altered in the above example, we could

substitute FAITH I-O for IDENT(PLACE) for a more specific analysis. However, the resulting

output would remain the same.

10.1.1.2

Markedness Constraints

Whereas faithfulness constraints operate to keep the input and output as closely related as

possible, markedness constraints aim to promote the most ‘natural’ forms in language,

giving rise to surface forms which are the least marked. Therefore, they can work in

direct opposition to the faithfulness constraints which will conspire to retain marked

features which are present in the input. By way of illustration, it has been observed that

the basic syllable structure is CV (Consonant + Vowel) and that, in some languages, a

coda consonant is disallowed during the parsing of syllable strings, making marked

structures such as CVC redundant. This can be achieved through the introduction of a

constraint such as *Coda:

*CODA: Parsed syllables must not contain a coda consonant.

This negative constraint which will aim to eradicate any coda consonant represented in

the input will then work in opposition to the faithfulness constraint Max, which guards

222

against the deletion of phonetic detail from the input form. An example of how this may

work in a language which does not allow coda consonants is given in 10.8

10.8

/CVC/ *CODA MAX

CVC *!

CV *

Using this ranking, the winning candidate is the unmarked CV form. However, a reversal

of this ranking would allow the coda to surface:

10.9

/CVC/ MAX *CODA

CVC *

CV *!

10.1.2

Optimality Theory and Language Variation

The increased application of constraint-based models such as OT), as opposed to rule-

based models, has gradually made the process of phonological transformation by rules

and derivation less common in phonological analyses. However, Anttila notes that “work

on phonological variation has continued largely independently of phonological theory,

often consciously emphasising its empirical character” (2002:214). Therefore, research

on the nature of phonological variation and change has largely focussed on empirical

sociolinguistic studies without the application of an accompanying phonological model.

However, the advent of OT has provided the linguist with an approach which can provide

novel strategies for modelling linguistic variation and change.

Some of these strategies are briefly outlined below.

223

10.1.2.1

Tied Violations

Hammond (1994) presented an analysis of stress placement in the Australian Aboriginal

language of Walmatjari. The stress placement in this language appears fixed in disyllabic

words, as initial-syllable stress is non-variable. However, stress on words with three or

four syllables has variable placement. Hammond proposes that this variability, grounded

in the observation that it is possible to group and construct metrical feet in more than one

way45

, can be shown as a tie between two candidates with respect to the number and type

of violations that they incur. The following tableau from Hammond (1994) illustrates

how candidates may be tied:

10.10

a b c d

1 x x x *

2 (x x x

3 x (x x

4 x x (x * *

5 (x (x x *

6 (x x (x *

7 x (x (x * *

8 (x (x (x * *

We can see that Candidates 2 and 3 both have no violations of the constraints a, b, c and

d. Therefore, they are equal when it comes to the constraint evaluation (see Hammond

(1994:5) for a detailed explanation of what these constraints represent). Thus, when

candidates are equal, with respect to their constraint violations, variability occurs between

them as surface forms. However, Anttila (2002:217) notes that this approach has two

inherent limitations. Firstly, the model is able to achieve ‘ties’ due to its representation of

only a subset of constraints which, he suggests, are designed to be universal and

45

For example, with the sequence ABC, metrical feet can be constructed from the left, as in (AB)C, or from

the right, as in A(BC)

224

sufficiently detailed enough to distinguish candidates, even if that means relying on much

lower ranked constraints present in the grammar. Secondly, one aspect of variation which

is not represented in this model is the degree to which each battling candidate surfaces,

and thus, it cannot account for one variant being dominant over the other in natural

speech.

10.1.2.2

Multiple Grammars

Analyses incorporating a multiple grammars approach propose that variation arises from

invariant grammars, whose individual winning candidates are presented as the choices for

the surface form. Therefore, the initial stages, when using this approach, are to determine

what the possible grammars are (that is, all the possible ways there are of ranking the

constraints in use) and what kinds of possible surface forms or trends these grammars will

produce. These predictions can then be tested against real data to see if the predicted

types of variation approximate the real speech data. Kiparsky states that “variation comes

from competition of grammatical systems (in the individual or in the community)”

(1993:2) and, using American English, applies this model to variation patterns in t/d

deletion. He notes that there are three ways of dealing with t/d clusters which lend

themselves to the deletion of final segments:

1) The word final t/d can remain as the coda (for example, [lost][everything])

2) The word final t/d is not affiliated with any syllabic position (for example,

[los]t[everything])

3) The word final t/d can be re-syllabified as an onset (for example,

[los][teverything])

225

The first option would lead to violations of syllabic constraints on complex codas, the

second would violate Align constraints as the syllable boundaries of the input would be

altered and, finally, option three would lead to the non-parsing and, thus, deletion of a

segment. The constraints proposed by Kipasky which reflect each of these options are

SYLLABLE-WELL FORMEDNESS46

, PARSE47

and ALIGN48

, respectively. Therefore,

three possible rankings representing three competing grammars can be proposed using

these constraints

Grammar 1 -

SYLLABLE – WELL FORMEDNESS >> ALIGN >> PARSE

Grammar 2 –

ALIGN >> SYLLABLE – WELL FORMEDNESS >> PARSE

GRAMMAR 3 -

SYLLABLE – WELL FORMEDNESS >> PARSE >> ALIGN

Each of these grammars would be activated so that candidates can be evaluated against

them in parallel and, as a result, variation would occur between the winning candidates of

each grammar.

46

This constraint ensures that any syllables in the output are formed according to the requirements of the

language. 47

The PARSE constraint, from containment theory (see Prince and Smolensky (1993, 2004) or McCarthy

(2008)) guards against deletion by making sure all phonological material in the input is parsed and

represented in the output. 48

ALIGN constraints ensure that constituents (such as syllable and morpheme boundaries) retain the same

relation to each other in the output as they do in the input (see Prince and Smolensky (1993, 2004)

226

However, Anttila (2002) notes that this model, while excluding several unattested t/d

deletion patterns, does persist with predictions of dialects which do not appear to exist.

These ‘improbable’ dialects exhibit, for example, “no quantitative difference [in deletion]

between the vowel and consonant environments” (Anttila 2002:224). However, Anttila

does note that once again, we cannot say that the model fails (as each tableau does not

fully represent the full set of constraints held within the grammar), and that we might

want to make a claim, for English, that there are some partial set rankings which would

remain inflexible in the ranking and thus restrict the scope of the available grammars.

10.1.2.3

Stochastic OT

Stochastic OT operates within one grammar. However, whereas classical Optimality

Theory, as outlined above, represents constraint ranking as a type of order relation (and

thus each constraint occupies a static position within the ranking order), Stochastic OT

allows for constraints to enjoy limited movement within the ranking hierarchy (see in

particular Boersma and Hayes (1999; 2001)). Therefore, each constraint projects a fixed

ranking value which establishes the initial constraint ranking in the grammar. For

example, in the graphic representation from Boersma and Hayes (1999:3) below, the

ranking is that of C1 >> C2 >> C3:

10.11

227

Variation within the grammar emerges because, in addition to the fixed ranking value,

each constraint is also allocated a random positive or negative value at the moment of

evaluation prior to its actual realisation. This resulting ranking value (or selection point)

generates the ranking of constraints as applicable at the time of that particular speech

event. Therefore, even though variability will occur, it is the initial setting of the fixed

ranking value which will determine which adjusted rankings will have a higher or lower

probability when compared to each other. Continuing the example of a constraint ranking

in 10.11 above, we can see how initial rankings can become compromised when the

random positive and negative values are added. The illustration in 10.12a shows how C1

and C2 are far enough apart that no overlap occurs (and therefore no variation in their

ranking results). However, 10.12b shows how C2 and C3 are close enough that, when

their random values are added, there is an overlap in the constraint ranges (meaning that a

variable ranking of C3 >> C2 is possible).

10.12a

10.12b

Both these examples are from Boersma and Hayes (1999:3)

228

It is the overlap in 10.12b which leads to variation if the constraints involved are those

which can influence the selection of a surface candidate.

However, since it is a requirement that each constraint employs the same standard

deviation, Anttila (2007:534) observes that there are some patterns that provide a

challenge for Stochastic OT. For example, he notes that, in a ranking of A B C, where A

must outrank B (A>>B) but C may appear in any ranked position49

, Stochastic OT would

not be able to maintain this grammar without either a reanalysis of data and subsequent

application of different constraints, or the provision for each constraint to have a different

standard deviation.

Even though these three approaches allow for variation of outputs, and have been used to

approximate the results of some established empirical data (as opposed to being

predictive), they do not allow for the evaluation of the underlying system when variation

leads to change. Instead, they are limited to recognising change as the re-ranking of

constraints to avoid tied candidates, the loss of an active parallel grammar or the

alteration of ranking values and random deviation values, which do not take into account

how these changes will affect underlying phonological structure.

In addition, even though these models have been applied to some empirical data, they do

not allow for any qualitative aspects of variation, such as style or intra-speaker

correlations within the variation pattern.

49

C>>A>>B or A>>C>>B or A>>B>>C

229

One possible approach that could model the variation and change in Mersea Island

English and represent some of the variation patterns highlighted by the data analyses is

Dispersion Theory. Therefore, the following sections will introduce the phonological

theory of Dispersion (as presented by Flemming (2004; 2006)), whose focus is on

underlying phonological structure and maintenance of structure. The main principles of

this theory will be discussed before attention is turned to how a theory of dispersion can

be incorporated into a three-tiered model of language in which the derivation and

variation of linguistic units is represented in both phonological derivation and

sociolinguistic practice.

10.2

Dispersion Theory

Flemming (2004; 2006), by means of phonetic classification and mapping of speech

sounds, aimed to explore, through an optimality theoretic framework, the “general

character of the constraints imposed on phonology by the need to minimise confusion”

(2004:232). By taking a more perceptual (as opposed to a productive) stance, the model

approaches a theory of speech by considering the conflict between constraints which

minimise effort on behalf of the speaker and those which ensure the perceiver can recover

enough distinctive information in order to avoid perceptual confusion. Thus, the theory

promotes the existence of constraints which maintain contrasts between individual units,

while the relative markedness of a sound is dependent upon the sounds with which it is

intended to contrast. Thus, contrasts are more marked the less distinct they are.

230

Using the distribution of a language’s vocalic elements, Flemming (2004:236) notes that

the selection of phonological contrasts within the vowel space is subject to three

functional goals:

1. Minimise Articulatory Effort – articulatory effort on the part of the speaker should

be minimised

2. Maximise Distinctiveness of Contrasts – individual forms must be as distinct from

each other as possible

3. Maximise the Number of Contrasts – the phonological inventory must contain as

many contrasts as possible

These fundamental goals each relate to a set of OT constraints which help to construct the

grammar. Thus, this type of functionalist approach influences the construction and

evaluation of the phonological system. However, McCarthy (2002:227) refers to

Dispersion Theory as a method which provides a type of meta-grammar whose

motivation is to evaluate candidate grammars as opposed to linguistic expression. This

contrasts with the classic ideals of OT in which it is the linguistic expressions (or units)

which are subjected to evaluations and not the linguistic generalisations. Therefore, the

basis of these two theoretical approaches to phonology cannot be directly compared (even

though they use the same basic machinery with respect to the application of constraints)

since their domains of application differ – whereas Dispersion Theory aims to provide

justification for the distribution of sounds in the underlying phonological system and not

to evaluate sound strings in preparation for realisation, Classic OT makes no assertions on

the underlying sound system but rather it proposes a number of surface form candidates

which compete for eventual realisation.

231

10.2.1

Dispersion Constraints

The conflicting nature of an inventory’s three functional goals (as outlined above) makes

OT an ideal framework since its very nature allows for the resolution of conflicts through

the relative ranking of representative constraints. Therefore, this section will introduce

the basic constraints involved in Dispersion Theory following Flemming (2004; 2006).

Minimise Articulatory Effort

Flemming suggests that “two of the fundamental forces shaping phonology are the need

to minimize effort on the part of the speaker and the need to minimize the likelihood of

confusion on the part of the listener” (2004:232). The articulatory effort on the part of the

speaker relates to the articulatory requirements of a particular sound in a particular

context. As a result, effort minimization on the part of the speaker may reduce the degree

of contrast between segments. For example, Flemming notes that “effort minimization

dictates that vowels should accommodate to the articulatory requirements of neighboring

consonants” (2004:235). As a result, a language will have a tendency to nasalize vowels

in between two nasal consonants (such as in man and sang). However, if the language in

question maintains a phonemic distinction between oral and nasal vowels, this process

will be minimized in order to maintain the contrast for perceptual purposes.

Minimum Distance – MIN DIST

The constraint MIN DIST ensures that the distinctiveness of contrasts is maximised. This

family of constraints make particular use of Formant structure as descriptors and phonetic

reference points. Formants are concentrations of acoustic energy which are combinations

of frequencies as air vibrates in the vocal tract. The three dominant formants in vowel

232

perception and description are Formants 1, 2 and 3 (F1, F2 and F3, respectively) - Formant

1 relates to the descriptor of vowel height, Formant 2 relates to the degree of frontness or

backness of a vowel and Formant 3 relates to the amount of lip-rounding present in a

vowel articulation (see Ladefoged (2006) for more information). Flemming (2004:238)

cites studies which provide evidence of F1 and F2 as the primary dimensions which

distinguish, and thus lead to the similarity and misperception of vowels. As a result, he

adopts the following “coarsely quantised three-dimensional vowel space” (2004:238).

10.13 A representation of vowel space according to Flemming (2004:238)

Using these matrices, vowels, or more particularly vowel space, can be referred to with

respect to their dimension values. For example, using only the first two formants, // and

/u/ can be described as F1 6, F2 4 and F1 1, F2 1.

The constraints which can be subsequently devised by using these classifications thus

take the form of:

Dimension : Distance

233

Therefore, the constraint MIN DIST F1:2 means that there must be a gap of at least two

squares between forms being evaluated on the F1 dimension. For example, the pair u ~

would be acceptable by this constraint but the pair u ~ would not be acceptable.

Since one requirement is that auditory distinctiveness should be maximised, less distinct

contrasts incur more serious violations. This is formalised by Flemming (2004:239) as

follows:

MIN DIST = D:n is ranked above MIN DIST = D:n +1

Therefore, by using this formula ranking we are saying that a violation of MIN DIST D:n is

more serious than a violation of MIN DIST D:n +1 since the former constraint will permit

a smaller distance between forms being evaluated which may compromise auditory

distinctiveness.

Maximise Contrasts – MAX CONT

Implemented as a positive constraint, MAX CONT aims to select an inventory of sounds

which represent the largest amount of contrasts. However, the requirements of MAX

CONT will conflict with those of MIN DIST and so, depending on their relative rankings,

MAX CONT will instead be allowed to select the largest viable inventory (that is, the

largest inventory permitted by the interaction and restrictions of other constraints). For

example, a language may guard against vowels which are too close together with respect

to F1 distance by ranking a MIN DIST constraint, such as MIN DIST F1:3, higher than MAX

CONT. This would ensure that contrasts are achieved up to, and including vowels three

spaces apart on the F1 dimension and any further attempt by the system at maximisation is

disallowed. This can be demonstrated by the selection of front vowels below. It is also

234

worth noting the use of ‘’ under the MAX CONT constraint. This is due to MAX CONT

being a positive scalar constraint. Therefore, each contrast between forms in a candidate

set is marked with a ‘’ since, as far as the constraint is concerned, the more ticks, the

better the candidate.

10.14

MIN DIST F1:3 MAX CONT

i - e - a

i - e - - a * * !

Even though the inventory /i - e - - a/ has a four-way distinction while /i - e – a/ has

only three contrasts, it loses out in the selection due to the violation of MIN DIST F1:3 by

the pairs e ~ and ~ a.

Once the inventory is evaluated, the phonological system then evaluates complete word

forms using strings of inventory forms that have been altered by certain phonetic and

contextual processes. It is at this stage that Flemming identifies a potential problem with

the theory, as the evaluation of contrasts between word forms must be cyclic (since

neighbouring word forms must, in theory, also be evaluated). However, he suggests that

an analysis using Dispersion Theory “avoids this problem by only evaluating inventories

of contrasting sounds (or short strings of sounds) in a particular context” (2004:269). For

example, a vowel inventory can be derived but in order to evaluate whether a specific

vowel + nasal is well-formed, we need to establish whether the nasal in question comes

from an inventory of consonants which are available to be selected in a postvocalic

position.

Building on these ideas, Flemming (2006) proposes and implements a tiered phonological

system consisting of three sub-components. He observes that, even though

235

distinctiveness constraints favour maximisation of perceptual differences between sounds

which are intended to contrast, these constraints would also prompt positional

neutralisation effects as results from processes such as assimilation. In other words, an

established contrast between forms in the phonological inventory would then be

neutralised in the surface realisation due to a lack of distinctiveness in a particular

phonological environment. However, the freedom to which distinctiveness constraints

may interact with other phonological constraints must be limited so that effects such as

contextual enhancement of contrasts (as opposed to contextual neutralisation), which are

not attested, cannot be imposed on the surface forms. Indeed, the acknowledgement that

distinctiveness is not the only mechanism required by perceptual devices, together with

the desire to keep any system of contrasts as consistent as possible across all contexts,

leads Flemming to the following:

“The distinctiveness constraints evaluate contrasts between

words only to check that the contrasts are adequately realised

on the surface – if not, they are neutralised (giving rise to

positional neutralisation). However, distinctiveness constraints

play no other role in the mapping from underlying to surface

form. That mapping is governed by constraints requiring

faithful realisation of the underlying contrasts in conflict with

markedness constraints (articulatory constraints, metrical

constraints, etc). The faithfulness constraints favour consistent

realisation of the inventory of contrasts in all contexts.”

(2006:1)

Thus, distinctiveness constraints apply to surface forms in context while any

enhancements of the system are carried out in the segmental inventory, free from context.

The result of this is a three-tiered phonological model in which particular constraints are

activated at specific times. Firstly, constraints which apply to the underlying system are

activated in the first tier, where only individual units, free from other linguistic contexts,

236

are present. The subsequent tiers then evaluate the underlying segments as they appear in

phonological strings (such as word forms) before further distinctiveness checks take place

in the final tier.

The following sections will outline the three sub-components of this model as presented

by Flemming (2006) before a subsequent discussion and demonstration of how it may be

applied to the variable data previously presented in this thesis. Accompanying this will

be highlights of any problems encountered in the application of this model to empirical

data, as well as how these problems may be overcome. The result will be a modified

phonological model which is designed to represent the interface of phonology and

sociolinguistics.

10.3

A Three- Part Phonology

Flemming (2006) outlines three sub-components which complement each other in an

individual yet hierarchical structure. Therefore, it is a stratal-type system through which

the derivational output of one stage provides the input for the next (see, for example,

Kiparsky (2000; 2008)). However, he notes that, even though these individual

components are distinguished, due to them enjoying limited interaction, “they are not

organised in a serial derivation” (2006:9), which is the case with normal stratal models.

These components are the Inventory, the Realisation and the Evaluation of Surface

Contrasts (ESC). This section will briefly outline each component in turn using

Flemming’s own examples, before demonstrating how they interact as a whole.

237

10.3.1

The Inventory50

“The Inventory selects a basic inventory of contrasting segment types” (2006:10)

To create an inventory of sounds is to select a set of individual forms which will be

consistent in their realisation irrespective of context. Therefore, a phonological inventory

will be made up of single, context-free sound segments, or phonemes. These will also

reflect the perceptual considerations of the linguistic system and effectively represent the

categorical divisions of perceptual space.

The selection of inventory segments relies upon the relative ranking of MIN DIST and

MAX CONT constraints (see 10.2.1 above) as well as what Flemming describes as

segment-internal articulatory markedness constraints (such as effort constraints).

Importantly, contextual markedness constraints are not activated by the system at this

stage.

By way of demonstration, consider the following selection of the vowel inventory of

Cantonese. Of particular interest here are the contrasts relating to backness and

roundedness. In addition, since the constraints rely on the classifications used in

Flemming’s perceptual grid, as presented in figure 10.13 above, it is reproduced here for

convenience.

50

From this point onwards, a capital letter (Inventory) will signify the underlying stage of derivation where

possible phonemic sets are evaluated. However, lower case (inventory) will refer to the actual set of

phonemes selected at this stage of derivation.

238

10.15 A representation of vowel space according to Flemming (2004:238)

10.16 From Flemming (2006:11)

Inventory:

MIN DIST =

F2:2 or F3:2

MAXIMISE

CONTRASTS

MIN DIST =

F2:3 or F3:3

MIN DIST =

F2:4

i u **! **

i y u * **

i y u *! *** *****

i y u *! *** *****

i u !

We can see that there is no input in the top left cell of the tableau. This is indicative of

the fact that candidate sets are not being evaluated against another underlying, or indeed

‘ideal’ inventory. Thus, it is the forms which are selected in the Inventory which provide

the input to the subsequent Realisation component (see below). The importance of not

activating contextual markedness constraints at this stage can now be understood – if

underlying vocalic contrasts are neutralised or altered in specific linguistic contexts, the

separation of contextual markedness constraints from the MIN DIST and MAX CONT

239

constraints allow for the relative stability of the underlying inventory across all contexts

(Flemming 2006:11).

10.3.2

The Realisation

“The Realisation component maps an input string onto its phonetic realisation”

(2006:11)

At this stage, individual inventory segments are put together in order to form segment

strings, or words. Thus, the input segments for this component are those which have been

generated by the Inventory constraints. Since formally individual and context-free

segments are now adjacent to each other, they have phonological context and, therefore,

contextual markedness constraints can now be activated. Flemming notes that it is these

constraints which motivate contextual variation, such as certain co-articulatory effects.

However, these alternations are not evaluated by MIN DIST. These constraints are only

reactivated in the third and last component which evaluates the output from Realisation.

Instead, the preservation of contrasts in Realisation is maintained by a mixture of

faithfulness constraints.

Therefore, the mechanics of this stage allow not only for contextual alternations, such as

co-articulation effects, to be established, but also aspects such as syllabification and stress

assignment to be calculated. This is due to the inclusion of certain phonetic information

which is accessed and mapped as the segmental strings are formed. The mapping of input

strings and evaluation of candidates at this stage resembles more of a Classic OT

approach with respect to the faithfulness constraints (aiming to preserve the qualities of

240

the input) being in direct opposition to markedness constraints (aiming to alter the input

in accordance with their specifications).

This can be illustrated by following Flemming’s example of Cantonese vowels. The

analysis models the neutralisation of front rounded and front unrounded vowels. Thus,

the contrast between front rounded and front unrounded vowels is neutralised following a

labial, as coarticulation renders a following unrounded vowel too similar to its rounded

counterpart (2006:12). Using the inventory established in 10.16 above, which provides

the Realisation phase with the possible vowels /i y u/, the tableau in 10.17 represents the

mapping of /pin/.

10.17 From Flemming (2006: 14)

Realisation:

/pin/ LABIAL

COARTICULATION

IDENT (F2) IDENT (F3)

pin *!

pin *

pn **! ***

N.B A superscript // denotes an unrounded labial constriction

By employing the specific constraint LABIAL COARTICULATION, this ensures that labiality

is spread to a vowel from an adjacent labial articulation. More specifically, Flemming

stipulates that it discourages any direct movement from fully closed to fully open lips

between segments which are adjacent. Thus, [pin] loses out due to the lack of labial

spread while, even though it has a minor violation of Ident on the F3 dimension, [pin]

succeeds in this evaluation. The result, therefore, is a direct conflict between input-output

correspondence constraints and a positional/ contextual coarticulatory constraint.

241

Once the realisation has been mapped and a candidate has been selected, each string will

then activate the final stage – The Evaluation of Surface Contrasts.

10.3.3

The Evaluation of Surface Contrasts - (ESC)

The ESC is where the distinctiveness of contrasts is assessed, based on the phonetically

mapped outputs of the Realisation stage. In order to do this, MinDist must once again be

activated in order to see if the outputs of Realisation are too similar to other potential

word forms. Importantly, however, Flemming notes that “there is no re-ranking of

constraints between components, although only certain classes of constraints apply in

each component” (2006:14). Therefore, the MinDist constraints must remain in the same

dominance relations in the ESC as they were in the Inventory.

The mechanics of the ESC allow for the target input to be evaluated in relation to a set of

other inputs with which it minimally contrasts. Ideally, this set would contain all possible

input forms which differ from the target by at least one segment as well as some

additional forms. Each input form from the set has the potential to be realised or indeed

neutralised with a neighbouring member of the set.

A merger can be obtained by having a violation of a higher ranked MinDist constraint in

relation to a constraint such as *MERGE (which prevents neutralisations). Therefore, if

the alterations made by contextual markedness constraints render, for example, a vowel

too close to a neighbouring Inventory vowel, the contrast between the two ceases to

surface. For example, in the case of Cantonese vowels, if the output [pin], from the

Realisation violates a highly ranked MinDist constraint, by virtue of being too close to

242

either /y/ or /u/ in the Inventory, the three-way contrast between /i y u/ will be lost in

favour of a two-way contrast. Conversely, if *MERGE outranked any MinDist constraint,

then the three-way distinction of /y i u/ would remain. Thus, with respect to Cantonese,

in which a neutralisation occurs, we can generate the following tableau.

10.18 From Flemming (2006: 14)

ESC:

/pyn1 pin2 pun3/ MINDIST = F2:2

OR F3:2

*MERGE MINDIST = F2:3

OR F3:3

MINDIST = F2:4

a.

/pyn1 pin2 pun3/

pyn1 pin2 pun3

*! * **

b.

/pin1 2 pun3/

pin1 2 pun3

*

c.

/pyn1 2 pun3/

pyn1 2 pun3

* *!

In this tableau, the first candidate set (a) shows how the distinctiveness of the modified

/pin/ compares with the other inventory vowels. However, candidate sets (b) and (c)

demonstrate mergers. Set (b) is a merger of /pyn/ and /pin/ in favour of the modified

[pin] form while set (c) is a merger of /pyn/ and /pin/ in favour of [pyn]. The contrast

between /y/ and /i/ is too small and hence set (a) is ruled out due to the violation of the

highest ranked MinDist constraint. This is also the case for set (c), since there is not

enough distance between /y/ and /u/ in order to satisfy MinDist = F2:4, and so, set (b) is

victorious.

The position of *Merge cannot be considered random, Flemming notes that the particular

constraint is likely to occupy the same place in the ranking hierarchy as Max Contrasts.

The subsequent implication is that “the minimum level of distinctiveness that is

acceptable for a contrast in the inventory should also be the threshold below which a

surface contrast is neutralised” (2006:14).

243

10.3.4

Evaluation of this Approach

The innovative incorporation of phonetic mapping in the Realisation allows for

phonetically detailed representations to be generated before further evaluation by

phonological mechanisms in the ESC. A benefit of this is, for example, that the

consistency of contrasting Inventory forms across all linguistic contexts is established

before they are modified and evaluated by the subsequent components. Another benefit is

that phonetically derived alternations can be established in the Realisation before being

passed onto the phonologically-based ESC, which then evaluates these outputs with

respect to the final distribution of contrast but cannot directly influence them. Thus, the

phonetics ~ phonology interface is well represented but, at the same time, limited due to

the restricted interaction between Realisation and ESC.

One drawback of this model, as it stands, is that it has been constructed by Flemming to

model and explore the nature of segmental neutralisations and positional enhancements

and not language or dialect variation and change. There are, however, aspects of this

approach which can benefit a variation analysis:

Consistency of the Inventory

Flemming (2004:232) notes that one desire of a phonology is that perceptual differences

between units should not be too subtle or else they are more likely to lead to perceptual

confusions. As a result, we can say that by allowing a maximally distinct inventory of

sounds, this reflects the natural tendencies and desires of the phonological system. Thus,

differences across language inventories will be established through contrasting rankings

244

of the MinDist and Max Contrasts constraints. This may even be extended to more

similar dialect varieties of languages, where we may want to claim similar underlying

inventories (reflecting membership of a particular ‘language’ or ‘language family’) but

contrasting derivations in the Realisation (reflecting variation within a ‘language’ or

‘language family’).

Contextual Variation

As we have seen, the Realisation phase takes into account linguistic context and can force

contextual deviations from the Inventory’s output selection. A benefit of this, with

respect to dialect variation, is the power to model differences between dialects of the

same language variety which adhere to particular linguistic patterns according to context,

such as Canadian Raising (as previously outlined).

Building on these observations, the following sections will present and discuss how some

of the patterns of variation observed in the Mersea Island English data (presented in

Chapters 4, 6 and 8 above) can be modelled using Flemming’s inventory-based model as

a guide. We will see that it is not only necessary to include the phonetic mapping of

underlying forms (which establishes linguistic context), but it is also necessary to

introduce a new component which can represent the interface between phonology and

sociolinguistics. This sociolinguistic component will be introduced as a way of

demonstrating not only contextual and stylistic variation, but also as a way of introducing

variation leading to dialect change over time.

245

10.4

An Inventory- Based Model of Mersea Island English

This section will explore how the empirical sociolinguistic data presented in this thesis

may be phonologically modelled.

Initially, I shall present examples of data following Flemming’s model (as outlined

above). However, once this has been done, it will be seen that, in order to represent the

data appropriately, modifications to the ESC must be made.

10.4.1

The Inventory

As previously discussed, the constraints which are active at this level are those which

regulate the underlying segmental inventory – MinDist, Max Contrast and any applicable

segmental articulatory markedness constraints.

In the previous examples and discussions, simple vowels have been discussed with

respect to the grid utilised by Flemming (2004; 2006). However, since the analysis of

Mersea Island English presented in this thesis centres on the variation of diphthongs,

certain descriptive modifications need to be made. Therefore, the following will apply as

outlined below:

MIN DIST – activated to regulate the spread of diphthongs across

articulatory and perceptual space.

MAXIMISE CONTRAST – aims to generate the largest inventory of

distinctive forms.

Segmental Articulatory Markedness Constraints – these constraints

would include those which aim to enhance the internal structure and integrity

of the diphthongs.

246

Traditionally, the diphthongal inventory of Standard English is made up of /a, a, e, ,

/ and three centering diphthongs /, , /, although these latter diphthongs tend to

surface as monophthongs (for example, there [] ~ [:]). These can be illustrated in

the vowel space as follows:

10.19a Illustration of Standard English Diphthongs

10.19b Illustration of Standard English centering diphthongs

As this inventory demonstrates, nucleic distinctiveness plays a greater role than offglide

variation – that is, there is greater flexibility in what can be a nucleus than what can be an

offglide. Therefore, we could claim, with respect to Standard English, that it is more

important, when selecting forms in the Inventory, to have contrasting diphthongal nuclei

247

than it is to have differing qualities of the offglide (/a/ ~ /a/ and // ~ // being the

only pairs to be constructed with contrastive offglides). Therefore, I would like to

propose the following Inventory constraints based on those previously outlined:

MIN DIST nucleus – In this case, minimal distance is referring to the diphthongal

nucleus/ onglide and it stipulates that diphthongal nuclei must not be in the same

or adjacent zones.

MIN DIST offglide – The constraint similar to that outlined above applies to the

diphthongal offglides.

MAX CONTRAST – The aim of Maximise Contrast is to exploit the articulatory/

perceptual space as much as possible.

DIPHTHONG CONTRAST (Internal) – This would be implemented as a segmental

markedness constraint. DC acts to maintain a particular distance between the two

diphthongal elements. Thus, focusing on height:

DIPH CONT HEIGHT 2 – diphthongal elements are separated by two zones with respect

to height

DIPH CONT HEIGHT 1 – diphthongal elements are separated by one zone with respect

to height

In addition, the segmental grid utilised in the following analysis will be simplified in

order to reflect the more familiar IPA division of vowel space:

10.20 A simplified segmentation of vowel space used for the analysis of MIE

248

Therefore, even though this grid does not reflect segments with the same phonetic

accuracy as that used by Flemming, the use of phonological target zones will help to

simplify the MinDist and Diph Contrast calculations.

Since this thesis is concerned with the behaviour and interaction of the three diphthongs

/a/, /a/ and //, this is where the following tableaux will focus. In addition, the tableaux

are not directly concerned, as in Flemming’s work, with the number of contrasts which

can be established through the optimisation of the vowel space. Instead, the primary

concern of the model is to examine how, by using dispersion theory, the integrity of the

diphthongal system is established and maintained. Thus, by using a subset of this system

(/a/, /a/, //), we can explore how the direction of change in Mersea Island English is

motivated and optimised, as opposed to seeing how the entire system is maintained over

various linguistic contexts.

To start, let us consider how we might derive a Standard English Inventory of sounds,

namely /a/, /a/ and //.

249

10.21 The Inventory tableau: To achieve the Standard English inventory.

PRICE ~ CHOICE ~ MOUTH MAX CONTRAST DIPH CONT HEIGHT-2 DIPH CONT HEIGHT-1 MIN DISTNUCLEUS MIN DISTOFFGLIDES

A *P *C *M ! *P~C *C~M *P~C

B *C *P *M *P~M *P~C

C ! *P *C *M *P~C *P~C

D *C *M ! *P *M *P~M *P~C

E *C *M ! *P *P~C

250

How to read this tableau - Explanatory Points

The top left cell is not representative of an input, since there are no inputs in the

Inventory. However, the lexical set references are included in this cell to indicate

which diphthong in the cells below correspond to which phoneme.

This is also reflected in the use of *P(rice), *C(hoice) and *M(outh) as violation

markers so that the reader can see which diphthong (or pair of diphthongs) is in

violation of which constraint.

The Max Contrasts constraint does not allow the same diphthong to be associated with

more than one lexical set. Therefore, candidate set C only has a two-way contrast

between // and // meaning that this variety, should it surface, would have a

underlying /a/~// merger.

Candidate set D still has a three-way distinction. However, /a/ has monophthongised

to /aa/.

The violations allocated in Diphthong Contrast relate to the differences between both

elements of the diphthong. For example, each candidate in set A violates

DiphContHEIGHT-2 as they only have a difference of 1 between the elements // and //, //

and // and // and //. However, only /aa/ violates DiphContHEIGHT-1, since the

diphthong has been monophthongised and thus there is no qualitative difference in the

vocalic articulation across timing tier slots.

The constraint MinDistNUCLEUS, which guards against contrasting nuclei occupying the

same or adjacent zones, is violated by, for example, the relation between // and //

(*price~choice) and // and // (*choice~mouth) in candidate set A.

When this is applied to the offglide, we can see that all candidate sets are in violation of

this constraint since they all replicate either // or // offglides within a set of proposed

inventory segments.

Therefore, what we can deduce from this tableau is that, the wider the elements are in terms

of articulatory and perceptual space, the more optimal the diphthong. This is an observation

which has been noted by those such as Yamada (1984:61, 67) who formalises this in her

investigation of diphthongisation as a result of the Great Vowel Shift under two

‘Dissimilative Diphthongisation’ rules. It is the first of these which is relevant to the present

discussion:

Dissimilative Diphthongisation (1) – Dissimilate the two Vs of a falling diphthong by

increasing the sonority of its first V.

251

As a result, the newly developed diphthong will continue to develop by lowering the nucleus

to a more sonorous position and thus creating a wider sonority (and articulatory) gap between

the two vowels.

Therefore, it seems reasonable to assume that the wider the diphthong, the less marked it is in

terms of perceptual distinctiveness (even though the articulatory effort involved will be

greater).

We can also present this claim by examining the nature of the quantum vowels /a i u/.

10.22

These vowels are generally proposed as being ‘unmarked’ due to their positioning and status

in phonological systems across languages. By extension, therefore, we might make the

assertion that the diphthongs /ai/ and /au/ are indeed ‘unmarked diphthongs’51 and that any

changes which occur to the phonological system are towards these more natural forms.

Indeed, we could claim that the winning inventory presented in 10.21 (which contains these

two diphthongs) reflects the inventory of the Younger speakers of Mersea Island English,

considering the analysis of their data.

51

It could also be claimed that /iu/, usually represented as /ju/ completes this diphthongal set. However, a

discussion of this feature is outside the scope of this thesis and will not be discussed here.

252

However, in order to establish the inventory of the traditional variants used by the Museum

speakers, alterations to the constraint rankings need to be made.

The primary difference between the Younger speakers’ /a, , a/ system and the Museum

speakers’ corresponding /, , / system is the (phonetic) distance between the diphthongal

nuclei and their offglides. Thus, the model needs to be able to represent differences between

the phonetic qualities of the nuclei, as this is where the linguistic change over time appears to

have been focussed. One solution, which is presented in 10.23, can be found in the re-

ranking of the Diphthong Contrast constraints.

Therefore, with less importance placed on the height contrast between the nucleus and the

offglide, a ranking reversal of DiphContHEIGHT-1 and DiphContHEIGHT-2 allows for the Museum

inventory to be the optimal candidate set.

253

10.23 The inventory tableau: To achieve the Traditional Mersea inventory.


A * P *C *M *P~C *C~M *P~C

B * P *M ! *C *P~M *P~C

C ! *P *C *M *P~C *P~C

D *P *M ! *C *M *P~M *P~C

E *P ! *C *M *P~C

254

If we consider the development of // and //, as discussed in Chapters 3 and 5,

respectively, we can see that the nuclei developments in this dialect from [] and [] to []

and [] (following the developments of the Great Vowel Shift) enhance the distance between

the diphthongal elements. Therefore, this change is in accordance with the optimisation of

the system outlined by Dispersion Theory. This optimisation then continues as the [] and

[] nuclei lower to [a].

However, this leads to speculation regarding the variable nature of the Older speakers’ data.

The data from this group represent variables from both inventory systems and may be

modelled using both variable rankings between the Diphthong Contrast constraints or by

proposing a multiple grammars type of analysis. This point will be examined in more detail

in 10.4.3 below once the analysis of the Museum and Younger speakers’ data has been

completed.

10.4.2

The Realisation of Contrasts

As previously outlined, this stage activates contextual markedness constraints and deactivates

Min Dist constraints. It is at this level that the individual forms selected by the Inventory are

mapped together as phonetic strings and therefore, certain types of information, such as

gestural articulation, become available for phonological constraint assessment. Since the Min

Dist constraints are redundant here, a ranking of correspondence (faith) constraints is needed,

as well as their opposing markedness constraints.

In this section, the apparent contextual effects of Canadian Raising, the lowering after

dorsals, smoothing of /a/ before /l/ and the preceding labial effect on //, which have been

255

highlighted in the data analysis above, will be modelled according to the Museum and

Younger speakers’ inventories.

10.4.2.1

Canadian Raising

As discussed above, Canadian Raising (CR) is the phonological process which sees the nuclei

of /a/ and /a/ realised as raised [~] and [], respectively, before voiceless consonants.

This can be achieved through a ranking of the following constraints:

RAISE VOWEL – No diphthongal nuclei of [+low] before a voiceless consonant

(However, [+low] required elsewhere)

IDENT1ST/ 2nd element – The specifications of the diphthongal elements in the input are

preserved in the output

The Raise Vowel constraint penalises the nuclei of diphthongs if they are low before a

voiceless consonant. In Standard English, the only diphthongs with phonologically low

nuclei are /a/ and /a/ and so these are the only diphthongs which would be affected by this

constraint. Working in opposition to this are the two faithfulness constraints Ident 1st element

and Ident 2nd element. These act to preserve the quality of the inventory forms which provide

the input for the Realisation stage. In order to achieve a CR correspondence, these

constraints can be ranked as follows:

RAISE VOWEL >> IDENT 1st element >> IDENT 2nd element

The two Ident faithfulness constraints have been ranked in this way to reflect the idea that it

is a worse violation to alter the quality of the diphthong nucleus than it is to alter the quality

of the offglide. In the tableaux below this ranking becomes apparent.

256

The Museum Derivation

The Inventory which was derived above with respect to the Museum data resulted in the

underlying diphthongs // and // for the PRICE and MOUTH vowels. Once these inputs are

passed to Realisation, they can be assessed in accordance with the activated constraints of the

level. Tableaux 10.24 and 10.25 demonstrate how a CR pattern can be derived using the

words house and houses and the constraints proposed above.

10.24

/hs/ RAISE VOWEL IDENT1st element IDENT2nd element

hs *! *

hs

haas *! * *

10.25

/h.zz/ RAISE VOWEL IDENT1st element IDENT2nd element

h.zz *

h. zz *!

haa. zz * *!

The faithful [hs] candidate is the winner in 10.24 due to its contextually appropriate raised

vowel. However, it is the losing candidate in 10.25 when [hzz] becomes optimal due to its

low nucleus and faithful offglide.

The Younger Derivation

The same constraints and relative ranking can be retained for the Younger speakers’

Realisation stage, even though the input from the Inventory differs from that of the Museum

generation. Unlike that of the Museum Inventory, the Younger speakers’ Inventory generates

the standard-like /a/ and /a/ diphthongs. The following tableaux demonstrate how, once

again, a CR pattern may be established.

257

10.26

/hs/ RAISE VOWEL IDENT1st element IDENT2nd element

hs *!

hs *

haas *! *

10.27

/h.zz/ RAISE VOWEL IDENT1st element IDENT2nd element

h.zz

h. zz *! *

haa. zz *

Here we see that in tableau 10.26, the faithful candidate is not considered optimal due to its

low vowel before the voiceless /s/ and therefore, even though it violates an Ident constraint,

[hs] is the winner.

10.4.2.2

The Dorsal Effect - A lowering of diphthong nuclei after /k/

The data from Mersea Island English (MIE) has consistently shown how a lowered nucleus of

both (a) and (a) variables correspond with a preceding dorsal. This lowering effect, which

could be considered as one of the promoters of the change from //~// to /a/~/a/, may be

classified as a contextual effect, brought about by phonological position of the diphthong.

Thus, any backness associated with the dorsal, may be translated and transferred to the

following vowel. However, if we propose that it is simply the backness which is transferred

to the following vowel, this would predict that a diphthong input such as // or // would

surface as, for example, [] or [], respectively. This does not occur in the data. Instead,

the main difference between the diphthong realisations following dorsals is in the dimension

of height. As a result, we might propose that a preceding dorsal is prompting the vocalic

feature [+low] to be adopted by the following diphthongal nucleus. If we are to consider this

258

as an effort saving device (as the tongue does not then have to move to a mid-front position),

this may also be strengthened by the nature of constraints which have been applied in order to

minimise articulatory effort on the part of the speaker. For example, Kirchner (2001; 2004)

proposes, through an effort-based approach to the process of lenition, a constraint named

LAZY which aims to minimise the speaker’s articulatory effort during the production of

speech. This type of constraint will, when applied to contextual co-articulation, logically

favour the process since the articulatory precision of phonological segments is compromised

in favour of articulatory fluidity.

However, since the primary constraints used for this model in Realisation are contextual

markedness ones, the constraint introduced here will be LOWER VOWEL, and this will thus

interact with the Ident constraints used in previous tableaux.

LOWER VOWEL – Any vowels following a dorsal consonant must represent the feature

[+low]

The terms of this constraint are general in that it is worded to apply to all dorsal consonants

preceding a vowel. Unfortunately, due to the nature of the real data recorded for MIE, as

well as a general lack of English words with //, // and // sequences, the proposal

here has included both dorsals in this constraint under the assumption that the voiced dorsal

would cause similar effects on the following vowel. However, it must also be acknowledged

that this may not be the case, since any //, // and // sequences are likely to be in

uncommon lexical items and, thus, they may behave differently due to their comparatively

lower frequency.

259

The Museum Derivation

Once again, the Museum Inventory will select // and // to be inputs. The following

tableau demonstrates how the word council may be derived:

10.28

/kn.sl/ LOWER VOWEL IDENT1st element IDENT2nd element

kn.sl *

kn.sl *!

kan.sl * *!

Thus, the faithful candidate loses out due to the highly ranked Lower Vowel constraint.

The Younger Speakers

As previously established, the Younger speakers’ input will be that of standard-like

realisations of the diphthongs // and //. Using the same ranking of constraints as used in

10.28, we can see that, this time, the faithful candidate does indeed win to become the surface

form and thus, the low vowel is maintained after the dorsal.

10.29

/kn.sl/ LOWER VOWEL IDENT1st element IDENT2nd element

kn.sl

kn.sl *! *

kan.sl *!

10.4.2.3

Smoothing - // to /:/ before coda /l/

The data presented in Chapter 6 demonstrated how the diphthong // was realised as a

backed monophthong before a coda /l/ in words such as while and mile. The data set also

showed that this variation occurred in the lexical item Island, despite the following /l/ being

260

an onset of the following syllable. However, the special case of Island will be discussed

separately below. Thus, the following tableaux are concerned with // followed by coda /l/.

There are two approaches to the OT representation of // monophthongisation before coda

/l/. The first would only be concerned with the quality of the vowel and not allow for

alternations in the quality of the [l] ~ [] allophones. The second approach would take this

alternation into account and thus not only model the vowel change but also the consonant

change.

10.4.2.3.1

Smoothing - Proposal 1

In order to obtain lengthening before /l/, we could propose the following constraints:

*//before cd /l/ - [] diphthongs are prohibited before coda /l/

*COMPLEX VOWEL before cd /l/ - only simple vowels may precede coda /l/

Though, on the surface, the first of these constraints may appear too specific, since it refers to

a particular diphthong rather than a class of sounds, it could be argued to be part of a wider

group of markedness constraints which construct the overall distribution patterns within our

phonology. For example, the // diphthong is also not found before //, and thus we could

propose that this is due to an inviolable constraint *//before // which does not allow //

forms to surface. This process of monophthongisation before /l/ is also not restricted to the

Mersea Island dialect, and so this constraint is applicable to other dialects of English (see, for

example, Britain (1991) for evidence of // and // monophthongisation before /l/).

The second of these constraints is used here to apply to //, although it may be applicable to

other segments, such as //. For example, we saw in Chapter 7 how lexical items with //

261

before coda /l/ lowered and monophthongised in some dialects due to language contact.

Thus, with further research, this constraint may be used to apply in more than the specific

domain of // smoothing.

The incorporation of these constraints, together with the Ident constraint, can be seen in the

tableaux below.

10.30 Smoothing (version 1) Museum Speakers – () becomes /:/ before coda /l/

/ml/ *//before cd /l/ *COMP VOWEL before cd /l/ IDENT1st element IDENT2nd element

ml *! * *

ml *!

m:l * *

10.31 Smoothing (version 1) Younger Speakers - /a/ becomes /:/ before coda /l/

/ml/ *//before cd /l/ *COMP VOWEL before cd /l/ IDENT1st element IDENT2nd element

ml *! *

ml *! *

m:l * *

The winning candidate for both groups of speakers is [m:l]. However, what the above

cannot seem to account for is a winning candidate with a different vowel quality, such as

[m:l]. A form like [m:l], which is the monophthongal form of /ml/, would, using the

constraints above, be tied with [m:l]. However, forms like this were not found to surface in

the data – only low, back monophthongs were found in this context. Therefore, we need to

consider other factors which might cause this trend.

262

10.4.2.3.2

Smoothing - Proposal 2

The noticeable aspect of the // - [:] alternation is the co-occurrence with coda /l/.

Generally, in dialects of English, when /l/ appears in the coda, it is the allophonic dark []

which is produced52

. Ordinarily, when a clear /l/ is produced, the articulation is one of a

coronal gesture followed by a dorsal gesture. However, in the case of dark [], the dorsal

gesture is articulated first (see, for example, Cruttenden (2001) for further articulatory

information). Therefore, we might propose that a type of regressive assimilation is taking

place as the articulators, while completing the preceding vowel, anticipate this dorsal gesture.

The result would be a lowering effect of the diphthong offglide and a lowering of the

diphthong nucleus.

Thus, following in the path of the set of SPREAD constraints (see Padgett (1995)), which aim

to represent feature spreading by extending a feature’s domain in the output, we could

propose the inclusion of a SPREADDorsal constraint which conspires to regressively lower and

back an adjacent vowel articulation.

SPREADDorsal – Any velar gesture of a dorsal sound segment is spread to the preceding

vocalic segment, if adjacent, resulting in the specification [+low, +back].

This constraint, together with those developed in Proposal 1, results in the following:

52

It is also common in modern dialects of English for the dark [] to vocalise (see Johnson and Britain (2007)).

It loses the tongue-tip contact and resembles a vocalic segment, although, since this vowel typically has a back

quality, it may be suggested to be in keeping with the analysis presented here. However, without a separate

analysis of /l/ vocalisation in the MIE data, no further assertions can be made at this stage.

263

10.32 Smoothing (version 2) Museum Speakers - () becomes /:/ before coda /l/

/ml/ *//before cd /l/ *COMP VOWEL before cd /l/ SPREAD Dorsal IDENT1st element IDENT2nd element

ml *! * * *

ml *! *

m:l * *

m:l *! *

10.33 Smoothing (version 2) Younger Speakers - /a/ becomes /:/ before coda /l/

/ml/ *//before cd /l/ *COMP VOWEL before cd /l/ SPREAD Dorsal IDENT1st element IDENT2nd element

ml *! * *

ml *! * *

m:l * *

m:l *! * *

Therefore, all the diphthongs are in violation of Spread due to their adjacent offglide while

[m:l] now loses out since it is a back mid vowel.

10.4.2.4

Island Monophthongisation

In contrast to the previous examples and discussions of Smoothing of // before coda /l/, the

monophthongisation of Island occurs before a syllable initial /l/ (or clear /l/). Therefore, we

cannot propose a spread of the dorsal gesture since it is the initial coronal gesture which is the

primary articulation for clear /l/.

It was discussed in Chapter 9 how the []~[] variants in the ‘Island’ lexical item do not seem

to be enregistered as a dialect marker of MIE. However, this word clearly patterns with the

coda /l/ words throughout the data sets. This suggests that the smoothing in ‘Island’ could be

either an over-extension of a phonological process (which has been restricted to this word), or

264

it is the residue of a more general phonological process now restricted to this item (possibly

through its frequency in the everyday speech of the locals).

Taking these points into account, we cannot therefore propose a set of constraints which will

induce smoothing in non-coda /l/ environments as that would lead to words such as ‘pilot’

and ‘highland’ being realised as [plt] and [hlnd], which never occur in the data.

Instead, by treating this as a type of lexical effect, we could propose that the Realisation level

allows for dialect specific constraints on lexical items to be activated, since all phonetic

information is available at this level and certain segment strings may be identified or

recognised. By allowing lexical effects at this level, it is implied that the

monophthongisation of ‘Island’ is not stipulated by the underlying phonology, and so the

Inventory is allowed to remain stable. Also, by not having the monophthongisation governed

by the sociolinguistic level, it reflects the consistent nature of the data and solidifies it as an

operative phonological process, which is then available for sociolinguistic manipulation (see

10.5 below).

Thus, we can introduce a constraint such as:

PRICE [] / ISLAND : The PRICE diphthong will monophthongise and back in the

lexical item ‘Island’

The use of the lexical set title PRICE in the constraint simply allows us to represent the

diphthong across each speaker group, as, for example, the Museum speakers and the Younger

speakers have different underlying Inventory forms for this diphthong. Even though the form

of this constraint is somewhat ad hoc, the application of it in the Realisation can be illustrated

265

for both the Museum speakers and the Younger speakers as follows in 10.34a and 10.34b,

respectively.

10.34a

The Museum Speakers

/.lnd/ PRICE [] / ISLAND IDENT1st element IDENT2nd element

.lnd *!

:.lnd *! *

.lnd *! *

:.lnd * *

10.34b


/.lnd/ PRICE [] / ISLAND IDENT1st element IDENT2nd element

.lnd *!

:.lnd *! *

.lnd *! *

:.lnd * *

Once again, the monophthongal variants are in violation of Ident2nd element as they are being

treated as a spreading of the diphthong nucleus to the timing tier slot of the offglide.

However, since this is the lowest ranked constraint, it does not prevent a monophthong from

surfacing. Indeed, it is a violation of the top-ranked lexical constraint which determines the

surface form. Thus, a violation of this constraint from any candidate which does not conform

to the criteria of the lexical conditioning constraint will mean it is instantly dismissed.

However, since the lexical item is stipulated as ‘Island’, it would only be an active constraint

266

when this phonetic form is provided as the input to the Realisation and not if other lexical

items (such as ‘pilot’) serve as the input. Another approach would be to suggest that the

Realisation phase does not contain types of lexical constraints, and that there is another

derivational level which activates at the lexical level. Therefore, the Realisation phase would

only hold constraints which are built around the phonetic information available at this stage,

while a lexical level would activate constraints according to any lexical considerations, such

as that relating to specific lexical items.

10.4.2.5

BOY

The data presented in Chapter 8 demonstrated how a preceding labial stop, notably /b/,

promoted more traditional centralised [] variants for the () variable in MIE. One

explanation advanced above for the preservation of [] following /b/ was a coarticulatory

labiality effect. Thus, the labiality features associated with the articulation of /b/ spread to

the following vowel causing a rounded vowel to surface.

This centralisation and rounding of // was also found to correlate with the lexeme BOY and

its derived forms (such as the plural form ‘boys’). It has also been noted (in Chapter 9) that

the lexical item ‘boy’ features in a prominent performance phrase ‘My boy’ and thus, this is a

possible factor in the maintenance of [] in this phonological position.

Therefore, two derivational options are available to us:

1) We can propose a type of SPREAD constraint to promote more rounded vowels

following a labial consonant (in this case /b/).

or

267

2) We can construct another lexical conditioning constraint which applies to the

lexeme BOY

The implications of these will be quite different with respect to our phonologies and so we

shall now consider each in turn.

A SPREAD constraint

If we proposed a constraint which stipulates that vowels are more rounded and centralised

following /b/, this would then apply not only to any /b/ + // sequence, but any /b/ + Vowel

sequence. The way of avoiding the latter scenario of constraint application would be to

specify that it is only the diphthong // (in the sequence /b/ + //) which is affected by the

constraint. For example, consider the following derivation which uses the Younger Inventory

segments as the input:

10.35

/b/ SPREAD LABIAL __

/b/ + // IDENT1st element IDENT2nd element

b *!

b *

However, this would still cause other /b/ + // sequences to be realised as [b]. For

example, consider the name ‘Boyd’ or the verb ‘boil’ in 10.36a and 10.36b, respectively.

10.36a

/bd/ SPREAD LABIAL __


bd *!

bd *

268

10.36b

/b/ SPREAD LABIAL __


b *!

b *

Even though other /b/ + // strings in the data are, in comparison to the BOY tokens,

particularly infrequent (only 2 tokens of preceding /b/ were not BOY derivatives – boiled and

boiler), there is no evidence in the data to suggest that other /b/ + // sequences favour the

rounded diphthongal nucleus53

. Therefore, even though this approach may initially seem the

most appealing in terms of OT theory, it does have drawbacks in its extended domains of

application.

A Lexically Conditioned Constraint

However, as arbitrary as the construction of lexically conditioned constraints may seem, the

construction of a constraint such as CHOICE [] / BOY eliminates the complexities of

constraints which will need to prompt labiality spread but limit its domain of application to

only a few individual lexical items. Therefore, we can define this constraint as follows:

CHOICE [] / BOY : The CHOICE diphthong will be realised as [] in the lexeme

BOY and any of its derived forms

The application of this constraint would work in a similar fashion to the lexical constraint

used for ‘Island’ (outlined above). Therefore, the tableaux in 10.37a and 10.37b demonstrate

its application with respect to both the inputs of the Museum speakers and the Younger

speakers, respectively.

53

‘Boiler’ and ‘Boiled’ were realised as [bl] and [bd], respectively.

269

10.37a

The Museum Speakers

/b/ CHOICE [] / BOY IDENT1st element IDENT2nd element

b

b *! *

10.37b


/b/ CHOICE [] / BOY IDENT1st element IDENT2nd element

b *!

b *

The exact composition of a constraint like this (as mentioned above) would be open to

interpretation and modification according to the nature of the underlying form to be changed.

Thus, we could not write a transformation constraint such as // [] / __ BOY and include

that in the grammar of the Museum speakers since, as established above, they are unlikely to

have // as the underlying form of the CHOICE diphthong. Instead, the constraint may be re-

written to accommodate any previous variants to which that group of speakers were exposed

(indicating a change situation) or simply to present it as // BOY (which would act as a

maintenance constraint for this variant in this specific context). However, once again, even

though lexically conditioned constraints seem the more practical approach to modelling the

variation of BOY morphemes, one area for future research would be to assess the validity of a

lexical level in the context of this model where constraints such as these are activated.

270

10.4.3

The Older Speakers – multiple grammars or variable ranking?

The transitional nature of the Older speakers’ data for each of the three variables under

investigation suggests that they generate what could be an extremely complex Inventory.

However, the nature of their data may be resolved by one of two means:

1. The activation of two Inventory grammars which, with variable selection, provide the

Realisation with different inputs (Multiple Grammars)

or

2. A variable ranking of the Diphthong Contrast constraints within the same underlying

grammar, or Inventory (Variable Ranking)

We have seen, while deriving the Realisation tableaux, that the ranking of constraints in this

phase is able to remain the same for both the grammars proposed for the Museum and

Younger generations. Thus, for the contextual patterns of variation observed in the data, the

constraint rankings may remain consistent. As a result, we can claim that the differences

between the Museum and Younger grammars are due to the selection of the underlying

inventory, rather than differences created by contrasting evaluations of phonetic forms. The

Older speakers’ grammar, therefore, needs to reflect access to both inventory sets if we are to

maintain that their data represent the intermediate stages of linguistic change – that is, a

change from the Museum Inventory to that of the Younger Inventory.

The multiple grammars approach would thus allow for two separate Inventory grammars to

exist in parallel. These would either feed into separate Realisation phases, or feed into a

singular Realisation phase. These can be illustrated as follows:

271

10.38 A comparison of underlying Grammars for the Older Mersea generation

The deciding factor would be whether we wanted to propose that the Realisation maintains

the same structure during linguistic change (Multiple Grammar 2). However, this appears

improbable as data relating to the sociolinguistic process of linguistic diffusion commonly

suggest that certain linguistic environments promote change more rapidly than others. This

kind of process could only be instigated by changes in contextual constraints which are

activated in Realisation. Therefore, even though the Realisation tableaux so far have

maintained consistent rankings, this may not necessarily be the case during periods of

linguistic change. The change would thus reach completion as fewer and fewer speakers,

over subsequent generations, activate both old and new inventories.

Conversely, the variable ranking approach would retain a singular underlying grammar, but

with a variable re-ranking of constraints which lead to the selection of different candidate sets

(and subsequently, different Realisation inputs). With respect to the dialect under

observation, this would mean a variable ranking between the two Diphthong Contrast

constraints, which determine the wideness of permissible diphthongs in the Inventory, and

produce the Inventories previously proposed in 10.21 and 10.23 above.

272

Therefore, it seems that the representation of Multiple Grammar 1 (above) is the more elegant

representation of the Older speakers’ grammar, since it allows for a separation of linguistic

systems which follow their own derivational structure. As a result, even though many

processes would overlap, an Inventory form generated in the Older speakers’ grammar would

not be subject to Realisation constraints generated by the new, incoming grammar.

10.4.4

Evaluation of Surface Constraint (ESC)

It is at this stage in Flemming’s model that the focus of evaluation turns back towards the

system as a whole and not only the individual ‘well-formedness’ of Realisation candidates.

Therefore, Min Dist constraints are once again required to assess how the modifications made

in the Realisation will impact on the distinctiveness of the system as a whole. In Flemming’s

example above, we saw that the vowel alternation which took place in Realisation resulted in

a form which was no longer distinct enough from other members of the Inventory set and so a

neutralisation resulted.

However, the available sociolinguistic data are not concerned with vowel mergers and so the

mechanisms which Flemming introduces to describe phonological neutralisation and

enhancements are not appropriate or adequate to model this data. This can be demonstrated

by the following example from a museum speaker’s use of a wider () variant following /k/.

273

10.39 A representation of a three tiered Museum speaker’s Grammar – lowering after /k/

INVENTORY


1 *P *C *M *P~C *C~M *P~C

*P *M ! *C *P~M *P~C

! *P *C *M *P~C *P~C

*P *M ! *C *M *P~M *P~C

*P ! *C *M *P~C

REALISATION

/k/ LOWER VOWEL IDENT1st element IDENT2nd element

k2 *

k1 *!

k * *!

ESC

PRICE ~ CHOICE ~ MOUTH

~ ~

*MERGE MIN DIST NUCLEUS MIN DIST OFFGLIDE

k ~ k ~ ka2 *P~C, *C~M *P~C

k ~ k ~ k1 *P~C, *C~M *P~C

274

How to read this tableau – Explanatory points

The winning candidate set in the Inventory is highlighted in bold so that its trace can be

seen in subsequent phases.

The winning candidate in the Realisation is highlighted by underlining so that its trace

can be identified in the following phase.

Thus, when it comes to the Evaluation of Surface Contrasts, we can see two sets

competing against each other – the first includes the modified form [ka], while the

second contains the original inventory form [k].

The *Merge constraint has been included, as Flemming notes that it is required to adopt

the position of Max Contrasts. However, it does not take effect as the newly modified

form does not merge with an existing inventory form. Therefore, it does not provoke a

violation from either candidate set.

As a result, both candidate sets share the same violations. Indeed, if only these phonological

constraints are active, both sets are tied in terms of violations incurred. At this stage we

could claim a tied candidate solution and use that to explain the surface variation which the

empirical data produced (see section 10.1.2.1 above). However, that does not seem a

satisfactory conclusion as there appears to be no mechanism for examining or explaining

which candidate may be picked for any one speech event. For example, studies have

proposed and demonstrated how, when more attention is paid to speech during style tests,

contextual co-articulation effects and variant usage changes (compared to that of speech in

more natural, relaxed contexts – see discussion below). These types of correlations in

sociolinguistic data cannot be accounted for through a purely phonological approach of tied

candidates.

Therefore, if we allow for linguistic contextual constraints to influence phonetic mappings

within our phonology, we must also, by logic, integrate mechanisms which have the power to

manipulate or override these constraints if necessary.

275

10.5

The Sociolinguistic Level of Speech Production

The following section will outline a proposal for an innovative third tier to be included in

the above model, which replaces the ESC and would lead to an integrated socio-

phonological model of variation. The sociolinguistic notion of Audience Design will be

explored alongside phonological theories of exemplars and frequency before attention is

turned to how aspects of these models can work together to provide the basis for an

integrated socio-phonological model.

10.5.1

Introduction

It has been demonstrated above that a faithful application of Flemming’s model is

inappropriate with respect to the sociolinguistic data presented in this study. That is

because, unlike Flemming, these data are reflecting variation and are not representative of

a phonological neutralisation or phonetic merger. Therefore, if the third tier was instead

governed by sociolinguistic constraints, this would give a speaker’s communicative

competence more recognition and the power to override certain linguistic constraints

which have been imposed at the previous level. For example, even though Tatham and

Morton note that co-articulation is not a phenomenon that can be switched on and off at

the will of the speaker, it can be described as “an involuntary process which, like many

others, can be voluntarily influenced”54

(2006:29). Thus, it is possible for a speaker to

limit or enhance co-articulatory effects during speech. In addition, there has been

extensive sociolinguistic research demonstrating style shifts in the speech of individuals

54

Original italics.

276

during interviews and speech production tasks. Thus, with respect to this model, the

Inventory provides the optimal set of sounds in the underlying grammar according to

specific distinctiveness constraints and these are maintained by the correspondence

constraints in the Realisation. However, the Realisation also introduces constraints

which modify the Inventory segments according to their phonetic mapping while the

Sociolinguistic level is able to modify and manipulate the phonological outputs of

Realisation. As a result, the Sociolinguistics level is where knowledge about how to

manipulate the phonology is collected and stored together with the social meaning of the

contending phonological forms.

10.5.2

Communicative Competence

Communicative Competence, a term introduced by and associated with Hymes (1972),

represents our ability to use our linguistic knowledge and our social knowledge together

in order to communicate effectively and appropriately. Thus, communicative

competence, by definition, must take into account any social factors which will influence

linguistic behaviour, such as the relationship between speaker and audience and the time

and place of the speech event. Indeed, Chambers (2002) notes that a speaker’s repertoire

of variants is usually linguistically based. However, in order for social interaction to

work “both the content of speech and its form must be appropriate to the speakers and

their interlocutors in the particular social context” (2002:11).

277

Through the adoption of a more descriptive anthropological approach, Hymes notes that,

when a person acquires linguistic knowledge, it is not only knowledge of what is

grammatical, but also knowledge of what is appropriate. For example, the person must

acquire competence in a wide repertoire of speech acts (see Searle (1979; 1999) for

example) and events, as well as the ability to evaluate others’ ability to demonstrate their

competence in communication. Hymes goes on to say that this communicative

competence “is integrated with attitudes, values, and motivations concerning language,

its features and uses” (2001:60). In addition, Hymes suggests that communication (as

well as language in general) reflects four assessment criteria. Therefore, in the context of

an utterance, this would be whether it is formally or linguistically possible, whether the

rendering of it is feasible in its implementation, whether it is appropriate to the context in

which it will feature and what the entailment of the utterance will be once it is performed.

The role of competence in communication is summed up by Hymes as “the ways in

which the systematically possible, the feasible and the appropriate are linked to produce

and interpret actually occurring cultural behaviour” (2001:67). The role of varying

cultural behaviour is also crucial to identifying and defining communicative competence

between societies since the nature of the acquired communicative competence will vary

according to particular societies and the social expectations as competency is shaped by

social experience.

The observation that people change how they speak has also been validated by empirical

sociolinguistic research. Labov demonstrated through his Attention to Speech model that

speakers changed their use of particular variants according to the type of speech that was

278

elicited from them. Thus, fewer vernacular variants were produced in the reading of

word-lists and minimal pairs, which were considered the most formal of contexts, since

“the speaker’s attention is focused directly on pronunciation” (Labov 1994:157), while a

higher percentage of vernacular variants were elicited in less formal contexts, such as

free-flowing sociolinguistic interview speech. Thus, Labov (and many others since)

demonstrated that the more self-conscious people were (that is, the more attention that

was paid to their speech), the less likely they were to produce vernacular variants. Once

again, therefore, this type of data shows how linguistic outputs may change in accordance

with non-linguistic factors.

This view that language use is a reflection of social structure can be said to be

encapsulated by Bell’s (1984:151) Style Axiom. This states that:

STYLE AXIOM – Variation on the style dimension within the speech of a single speaker

derives from and echoes the variation which exists between the speakers on the “social”

dimension (1984:151)

Thus, Bell demonstrates, through a number of studies, how the range of variation along

the style dimension is always smaller than the range along the dimension of social status.

Bell’s Audience Design Framework thus allows a speaker to draw upon the linguistic

resources which are available within their community. Therefore, as Bell (2001:145)

suggests, the speakers are not passive participants. Instead, they are active in their

exploitation of the linguistic resources at their disposal. In addition, the model allows for

speech modification to occur in response to an absent audience – referee design. Referee

Design, as modified and described by Bell (2001), is “the linguistic expression of

identification with a reference group who are important to the speaker” (2001:163).

279

Therefore, even though representatives of such a group may be absent from the direct

audience, the linguistic forms associated with that group may be projected by a speaker in

order to demonstrate certain social or cultural affiliations.

Building on Bell’s original 1984 work, and the idea of numerous social factors

influencing linguistic choice, Preston (1991) investigated the role of status on speech

through numerous statistical analyses of sociolinguistic data. He explains that, in this

context, status “refers to a number of demographic features having to do with social class,

age, sex and so on” (1991:38) and develops the Status Axiom:

STATUS AXIOM: Variation on the “status” dimension derives from and echoes the

variation which exists within the “linguistic” dimension (1991:36)

Thus, the range of statistical probabilities relating to status will be contained within the

range of probabilities created by some linguistic factor. As a result, Preston presents his

status axiom as a stronger version of Bell’s earlier style axiom. When this is related to

sociolinguistic variation, Preston (1991:52) implies a hierarchy of importance when

linguistic forms are selected:

Strongest Linguistic Environment - Representing universal and language-

specific conditions on features

Mid Status - Represents more or less permanent aspects

of identity

Weakest Style - Controls (through the identity of the register)

only the selection of a point along a

continuum

280

When these approaches and points are taken as a collective, it is indicative that, as a

speaker, we can alter our linguistic output in response to ever changing social contexts, as

well as actively initiate changes in our linguistic forms as a way of dynamically

projecting our self-image. However, if we are to follow Preston’s hierarchical

suggestion, aspects such as the speaker’s own status characteristics will take priority over

style factors in the selection of variants, while linguistic considerations and constraints

will take overall priority.

10.5.3

Where does that leave phonology? Information through exemplars

If the underlying notion that speakers can alter their language outputs in accordance with

social and contextual factors is correct, it is logical, by extension, to propose that

linguistic forms are somehow encoded with social information which enables the speaker

to select appropriately for any given speech event.

The introduction of exemplar theory into phonology (for example, Pierrehumbert (2001))

proposed that detailed phonetic memories of individual linguistic forms can be stored.

These mental representations of phonological targets and patterns are thus added to and

modified each time they are perceived. This leads to the construction of exemplar

‘clouds’. Within these clouds, those exemplars which are activated more frequently are,

as a result, able to be retrieved more quickly when the time comes for production. This

model is therefore able to reflect the behaviour of speaker accommodation since “speech

patterns, which are heard recently and frequently, dominate the set of exemplars for any

281

given label, and therefore guide the typical productions” (Pierrehumbert 2001:13). In

addition, although Pierrehumbert does not attempt to model social and stylistic factors

within the exemplar model, she does suggest that these ‘deeper causes’ which influence

exemplar selection may allow for selection of different parts of the exemplar cloud in

different situations (2001:7). Therefore, even though it is not explicitly dealt with, once

again, this assertion suggests that there is a type of sociolinguistic knowledge attached to

linguistic forms, because, if there is not, we need to ask how these deeper causes might

select from less active members of the exemplar cloud appropriately and with accuracy.

Indeed, the answer to this question might be found in the idea that, when we perceive

linguistic forms, we not only decode them on a linguistic basis, but also on a social basis.

For example, consider the following phrase:

he’s right [: ]

Even though it is a syntactically and morphologically standard construction, the

phonological form is non-standard English – initial /h/ is dropped, raised /a/ nucleus,

word-final glottal stop. All three of these features, once perceived, can be coded and

stored linguistically. However, what if we add some extra information about the context

in which it was perceived:

282

he’s right [: ] - Speaker Close friend ‘X’

Male

30 years old

Colleague

Cricket fan

- Context

One-on-one chat in ‘X’s

living room

No one else present in room

No one else present in house

- Topic Cricket

Once this information is acknowledged and coded, the linguistic features can be stored

with these social associations. Therefore, [] would be associated with relaxed

conversational exchanges with ‘X’ and a sporting topic.

Now let us consider the same utterance produced by ‘X’ in the same context, but, instead

of a sporting topic, the topic relates to that of a professional, work-related issue.

he’s right [h: ] - Speaker Close friend ‘X’

Male

30 years old

Colleague

Cricket fan

- Context

One-on-one chat in ‘X’s

living room

No one else present in room

No one else present in house

- Topic Work conflict

283

Two changes have been made to the form of the utterance – the initial /h/ of the pronoun

has been retained and the pronounciation of the diphthong has widened so that it is

produced more like the standard form. The associations with ‘X’ as an interlocutor and

the associations with the contextual situation remain the same, but new topic associations

have been established.

By establishing numerous associations between linguistic forms and, for example,

interlocutors, social contexts and topics, the perceiver continually gains experience of

what to expect from each factor. Thus, a speaker can then use this information when it

comes to their own production of appropriate, or indeed inappropriate, linguistic

selection. Therefore, accommodation is possible (not just to a speaker or group of

speakers), but also accommodation to, for example, a situation or topic is possible. As a

result, speakers are able to switch between speech styles and registers dynamically as the

speech event or conversational exchange unfolds in time.

This type of awareness of social factors through linguistic associations may also account

for speaker-initiated variation such as divergence. If a speaker wishes to adopt an image

or persona of a group which holds significant appeal (such as that which Bell discusses

under referee design, see above), the associations can be made through perceptions of

experience (whether direct or indirect) and then linguistic forms can be manipulated

accordingly.

284

However, if we are to follow Preston’s hierarchy, outlined above, certain associations

will take priority over others when it comes to the selection of linguistic forms. Thus,

linguistic forms will be primarily considered on linguistic criteria, such as language

specific criteria. The selection will then be biased towards associations of ‘status’ and,

lastly, surface forms will be chosen according to associations of ‘style’.

10.5.4

How this can be implemented through the three-tiered model

The acknowledgement of linguistic factors as the primary influence on the derivation or

nature of linguistic variants could be said to reflect the first two tiers of the model as

previously discussed (i.e. the Inventory and the Realisation). In the context of this

analysis, language or dialect-specific conditions on diphthong form shape the underlying

inventory and provide the phonemic entries. Then, in the Realisation, additional factors,

such as co-articulatory and contextual effects, are considered as the result of phonetic

mappings of elements in phonemic strings. Therefore, these linguistic phases have

primacy in the shaping of underlying forms before they are produced. However, as

discussed above, surface forms may not always behave in a linguistically consistent way

as variation occurs in, for example, contrasting stylistic and social contexts, and co-

articulatory effects may be enhanced or reduced. In other words, to be truly linguistically

consistent would result in the production of the underlying inventory forms in every

linguistic context. However, allophonic alternations are prompted in the Realisation

which allows for linguistically motivated variation and extra-linguistic factors can lead to

variation between these allophonic structures. Therefore, we need to construct a

285

component which allows the speaker to retain trace memories of the output from both the

Realisation and the Inventory so they too can be evaluated according to real-time

sociolinguistic considerations.

Using the notions of status (this time on the part of the speaker), style and context, we

may construct something like the following matrix:

10.40

Audience Status e.g. who? Style e.g. How? Context e.g. where?

The knowledge about each category is gained from a person’s continuous linguistic

experience. For example, only by experiencing certain speech contexts, either directly or

indirectly (such as speaking in a job interview or a sermon in Church), can a speaker

collect information about which linguistic form is appropriate when. By way of

illustration, consider the lowering of diphthong nuclei after dorsals for a Museum Mersea

Island speaker. Their Inventory produced // while the Realisation generated []. Here

is how we might evaluate both according to context and style:

286

10.41

Style Context

Formal Informal Formal Informal

By using tick marks ‘’, the characteristics of each candidate in terms of social and

stylistic value are mapped. Therefore, rather than acting as a ‘selection through

violation’ process, as in classic OT, the ticks represent the extra-linguistic mapping of,

for example, the social and stylistic attributes of a linguistic form onto a matrix of

categories. As a result, the example above shows how a speaker has categorised [] as

being associated with informal, relaxed speech styles and familiar contexts, while the

reverse is true of []. Furthermore, we can see how these would potentially interact as a

formal unfamiliar situation would prompt the selection of [] while an informal familiar

situation would prompt []. However, this type of binary classification seems

methodologically crude in that we are unlikely to make such clear-cut distinctions for

each interaction in which we are involved. For example, a speaker would find a conflict

in variant selection above if the speech event represented a formal style in an informal

context, such as the reading of a book passage at a friendly, local book club held in a

member’s house. Therefore, we might propose a more individual or tailor-made division

which reflects our personal situation and experience. By way of illustration, as a 28 year

old, PhD researcher, lecturer and speaker of Mersea Island English, I might include the

following in my sociolinguistics matrix:

287

10.42a

Style

Chatting Lecturing Telephone Interview Performance etc.

10.42b

Context

Home Classroom Supervisor’s

Office

Restaurant Friend’s

House

etc.

Thus, the further a classification is to the left of a category, the more salient it is to me.

The inclusion of ‘Performance’ under Style allows for me to perform, or enhance certain

linguistic characteristics. For example, it was mentioned above that certain traditional

dialect features may become entrenched in certain set phrases. These performance

phrases may not be in keeping with the rest of a person’s speech pattern, but they can be

accessed during a speech event if necessary. This could also include aspects such as

mimicry of accents which are not socially related to the speaker – for example, the

performance of an American accent during story telling. In addition, I have advanced

many classifications under ‘Context’ that pertain to working at the University where,

depending on the season, I spend most of my time. This will obviously not be the case

for others who hold different jobs or indeed even for those who partake in similar

activities to my own. Also, allowing for this much detail leads to a particular linguistic

form being associated with more than one cell on the sociolinguistics matrix. However,

if I do not experience any of these classifications for an extended period of time (for

288

example, I move to a different geographic region or change jobs), then I will gradually

lose traces of previous associations and acquire new ones according to my new

experiences. With these new experiences may come new variant exposure, which the

phonological and sociolinguistic system will need to decode and re-classify.

It is a significant implication of this proposal that, since both the Inventory output and the

Realisation output are evaluated at this final stage, the former must have a trace which

can then be retrieved by the production system. However, the model must also be able to

account for surface variants which do not seem to be represented by either the Inventory

or the Realisation outputs.

If we return to Exemplar theory, it was mentioned above that the forms we perceive are

stored in exemplar clouds from which the speaker can pick the strongest when it is time

for production. To accommodate this notion in the current model, it can be proposed that

the perceived exemplar forms in a particular situation are stored before they too present

themselves for evaluation alongside the Realisation and Inventory outputs. Thus, if I, as

a perceiver who has // as an Inventory form, consistently hear at work [] forms used

in all linguistic contexts (not only in those where I would expect lowering), I would

associate that form, in all linguistic contexts with work and the people associated with

that institution. Thus, I would evaluate, for example, [kt] and [st] as permissible and

valid forms in a work context. Then, if I chose to accommodate to this pattern, I too

would use [] in both and not just after the dorsal.

289

This also gives an important role to the people with whom we interact. It is these people

who provide us with the exemplar knowledge and who each have the potential to be an

audience themselves. As a result, our interlocutors provide a great influence over how

we design our speech and must be included in our evaluation of surface candidates. As

we grow up and learn the social rules of language interaction and use, we are

linguistically groomed with respect to how we can say what to whom (leading to much

anecdotal data of parents correcting their children’s pronunciation or apologising for their

children’s often embarrassing breach of interactional conventions).

Once again, therefore, it would not be methodologically appropriate to suggest that we

categorise our interlocutors in a binary way (such as known vs. unknown), but rather as

individuals or, at the very most, a group. The former would allow us to categorise and

respond linguistically to the behaviour of another while a group affiliation (such as that

reported in communities of practice studies - Eckert (1989; 2000), Mendoza-Denton

(2008) and Meyerhoff (2002)) would prompt linguistic behaviour as expected from the

group norm of affiliated members. Indeed, this type of group audience could also be

activated in order to signal membership, even when the speaker is removed from the

vicinity of the other members, in accordance with Bell’s referee design (as previously

discussed). Taking these observations into account, the ‘Sociolinguistic Speaker Matrix’

may be represented generally as follows:

290

10.43

Audience (Direct or Indirect)

Individual A Individual B Individual C Individual D Group 1 Group2

The three aspects of Audience, Style and Context would, therefore, all work together in

the evaluation of potential surface forms.

10.44

Audience Style Context

A1 A2 A3 S1 S2 S3 C1 C2 C3

However, it must be noted at this point that there is a potential clash of terminology. In

the above proposal, the terms ‘style’ is referring to what type of speech is appropriate for

the speech event and, therefore, answers the question How should I speak in situation X?

As a result, this is more in keeping with more traditional Labovian interpretations of style

(such as that used by early sociolinguistic methodologies which would elicit and compare

speech from word lists, reading passages and casual speech). However, a strict adherence

to Bell’s model would lead to the interpretation that ‘style’ is in fact the ‘audience’ and,

therefore, these two should not be separated into individual analytical entities.

291

Therefore, in order to avoid confusion, the term ‘Speech Type’ will be used in place of

‘style’. The incorporation of this term allows for the maintenance of a three-part system

between who is being addressed (Audience), how the speech event is altered to conform

to stylistic factors (Speech Type) and the setting or where the utterance takes place

(Context).

10.5.5

The Activation of the Sociolinguistics Matrix and Evaluative Parameters

The inherent nature of any natural (non-scripted) interaction is that it is an ever-changing,

dynamic process. Even the most mundane exchanges do not remain static with respect

to, for example, audience, topic and context. By way of illustration, a chance meeting

with a friend on the street may include many rapid changes in topic, while a group

interaction may include many changes in the direct and indirect audiences. In addition,

the need for emphasis, and thus clearer articulation may arise, or the need to moderate, or

alter one’s speech in accordance with perceived stereotypes may be desired.

Therefore, any model of speech must allow for the ever-changing dynamics of social

interaction and the influences this will have on the linguistic system. Focussing on

phonology, however, the phonological system may provide the speaker with the optimal

options for production, in accordance with underlying structure, but there must be a

flexible mechanism in place so that the demands of communicative competence can

override the phonological system’s recommendations if necessary. In other words, the

phonological system provides, stores and maintains a number of possible linguistic

292

options while the range of interacting sociolinguistic factors determine which one is

selected. However, the range of linguistic options will be limited by the linguistic

system, in this case the phonological system, since it may only generate options which

are phonologically well-formed and appropriate to the speaker’s linguistic system.

This can be graphically represented by column activation on the sociolinguistics matrix.

Consider the following interview scenario and subsequent, simplified, matrices:

A and B, who are Older Mersea speakers, are husband and wife. They are being

interviewed by J, a Younger Mersea speaker, in the living room of their house. In the

middle of talking to J, the phone rings outside the interview room and A goes to answer.

The caller, it is later revealed, is A’s son X.

10.45a A’s Sociolinguistic Matrix for ‘might’ – Addressing the interviewer J in the

presence of his wife B

Audience Speech Type Context

B J Informal Formal Familiar Unfamiliar

mt Inventory

mt Realisation

mat Interviewer

10.45b A’s Sociolinguistic Matrix for ‘might’ – After answering the phone to X in a

separate room

Audience Speech Type Context

X B J Informal Formal Familiar Unfamiliar

mt Inventory

mt Realisation

mat Interviewer

293

The subscripts in the far left column indicate the source of the evaluated candidates or

variants. Note how the interviewer’s form acts as an input at this stage since that is the

one being perceived, decoded and stored in association with the relevant social factors.

This allows for new phonetic forms to be interpreted by the phonological system and

evaluated accordingly by the sociolinguistic system so that accommodation to the new

form is possible.

The highlighted columns indicate which social aspects are relevant at the time of the

speech event. Therefore, the context does not change, meaning that both utterances were

carried out in a similarly familiar location, in this case, A’s home. However, both the

interlocutor and the style change, the former due to the shift in addressee, and the latter

due to the break in conversational structure (the phone conversation is not viewed by the

speaker as being part of the recorded interview structure, but rather a separate casual

exchange). Therefore, the parameters for candidate evaluation also change.

In 10.45a, the speaker’s choice is between [], which the interviewer is using and which

is also associated with the current speech style, or [], whose only relevant associations

(in this simplified illustration) are to do with the familiarity of the context. Conversely,

when the audience and style of speech changes, there are no associations of [] in the

activated cells, making [] the likely surface form and accommodation to the interviewer

less likely, even though J maintains the type of status associated with over hearers.

294

The benefit of this approach is that it is able to illustrate surface variation in a dynamic

conversational structure by highlighting what possible factors may have altered as

different variants are used. Indeed, the rather simplified examples above can be adapted

to reflect any number of additional or personalised factors. For example, the only

additional input in the Sociolinguistic Level was from the interviewer. However, in a

group situation or dialect contact situation, there may be a number of new forms added

from speakers, which will work to bolster their respective exemplar clouds and create

numerous new sociolinguistic associations. If the contact with these new forms is

temporary, the associations will remain weak. Conversely, if these associations are

continually activated, they will strengthen as they are used with increased frequency.

Eventually, these strengthened forms will cause a shift or change on the part of the

speaker.

The following section will now discuss how language change may be represented and

reflected by this three-tiered model.

10.6

The Representation of Types of Change

The model outlined above has so far demonstrated how underlying phonological forms

can be generated to provide a structured and optimal phonemic inventory. The members

of this inventory are then passed onto a secondary level which activates and imposes

certain phonetic constraints on their production once placed in certain linguistic contexts.

These two stages, governed by linguistic factors, provide the basis and inputs for the final

295

stage of sociolinguistic evaluation. It is this final stage where the interaction of

phonology and sociolinguistics can be seen and surface variation between linguistic

forms can be illustrated according to various sociolinguistic (or external) demands.

However, it seems that any model of variation would not be complete without the

provision for representing language or dialect change. Chapter 9 has summarised the

characteristics and types of change found in MIE as well as providing explanations for

some of the lexical variation (namely that associated with ‘Island’ and BOY).

Therefore, the following sections will illustrate how we can incorporate and describe the

general theory of language change into the three-tiered model presented so far in this

Chapter.

Minkova and Stockwell (2003), investigating a number of vowel and diphthong shifts,

state that “once phonetically motivated surface changes gather enough momentum, they

percolate into the underlying structure, creating new phonologies” (2003:184).

Therefore, in the early stages of change, contrasts in the underlying structure of the

Inventory would remain unaltered. This can be reflected in both the data and the model

presented so far. Applying the principles of this quote from Minkova and Stockwell, we

can say that ‘phonetically motivated’ changes are those natural changes which arise in the

Realisation stage. With respect to the MIE data, an example of this could be the lowering

of diphthong nuclei after dorsals. This change does not affect the underlying structure of

the Inventory as, in this model, these two stages (Inventory and Realisation) are separate.

296

However, even though the initial domains of application are phonetically restricted (for

example, diphthongal nuclei are only lowered after dorsals), once these patterns are

established, frequently used and reinforced by the speaker and speech community, other

domains of application may emerge until the Realisation stage is altering all or nearly all

inputs from the Inventory. This type of change which allows for certain phonological

environments to change independently of other phonological environments reflects the

principles of Neogrammarian sound changes (see, for example, Labov (1994)). Once the

momentum of the change has gathered, the change may be filtered up to the inventory

and the underlying structure will be reassessed and altered. An immediate advantage

would be that processing in the Realisation would be reduced as less context-specific

constraints would be activated to apply to specific inputs. It is at this point that the

Inventory might need to readjust in order to conform with the constraints active at this

level. In the case of this model, these would be the dispersion constraints which maintain

the structure of the phonemic inventory.

By way of illustration, consider the changes in the PRICE vowel from [] to [] in MIE55

.

The data have shown that this diphthong is subject to processes associated with Canadian

Raising, lowering after dorsals and smoothing before coda /l/. The following represents

these processes with respect to a Museum speaker:

55

Let it be noted that by selecting this variant, I am not proposing that this is where the change in MIE

originated. It is being used only for illustrative purposes.

297

Canadian Raising - [] before voiceless consonants

[] in all other contexts

Dorsal Lowering - [] in all other contexts

[] after dorsals

Smoothing - [] in all other contexts

[:] before coda /l/

All these processes are derived from the context-specific constraints in the Realisation.

Therefore, to minimise the processing load, it would be advantageous to simplify these

constraints where possible. It is possible to suggest that dorsal lowering may be extended

to other preceding articulations. However, without any convincing articulatory argument

(provided by the data) to accompany this suggestion, together with preceding dorsals

being infrequent in the data, it seems unfavourable to suggest this as the main driving

force behind the change to []. Conversely, it was suggested in the previous chapter

that the Canadian Raising type patterns found in the data did not appear to be entrenched

(or enregistered) as a phonological rule. It was suggested that this helps provide support

for it being a natural pattern of reallocation as a product of change through contact, as

opposed to a natural pattern which has become a linguistic rule. The result is that the

wider diphthong has a more diverse range of application with respect to linguistic

contexts and lexical items. Over time, the increased frequency of [] usage will then

overcome [] in the perception of others (such as the next generation of speakers) and

result in [] being the majority variant which has the more complex set of phonetic

constraints in the Realisation. This would subsequently put pressure on the Inventory to

adopt [] as an underlying form //, which, it has been proposed, has taken place and

been completed in the Younger generation speakers. The Younger speakers, by this

298

point, do not have the complex set of constraints in the Realisation, with the exception of

the ‘Smoothing’ constraints and the now raising of [] to [] before voiceless

consonants (which reflect the Canadian Raising rule). However, the latter seems to be

almost lost in the speech of the younger generation and so even this process may not be

present in some speakers. With respect to smoothing, one explanation for its preservation

in the younger generation may be that the resulting sound from the Realisation is [:]

(from the input //) and this is neither of the [] or [] forms which are competing for

Inventory representation. However, as the data in Chapter 6 suggest, this smoothing is

also on the decline in the Younger speakers and so it is possible that this contextual

constraint is already on the way to being lost from the grammar.

The change in the Inventory structure between Museum speakers and Younger speakers

is also bolstered by the behaviour of the change from [] to [] in the MOUTH vowel.

With the exception of Smoothing, the MOUTH vowel demonstrates the same contextual

variation as the PRICE vowel and, therefore, it could be said that the Inventory is re-

structuring towards more optimal diphthongs (with respect to their internal structure and

the structure of the diphthong system overall). Therefore, they meet with less resistance

as both are available for alteration.

The following flow chart demonstrates the progression of these changes:

299

However, this approach is primarily focussed on linguistic factors (namely naturalness)

as the main driving force of change with sociolinguistic factors there to reinforce and

promote the newly generated variants. Therefore, since the Realisation forms contribute

to the input of the Sociolinguistic level, we need to allow for driving forces of change

which are outside the domain of phonetic motivation.

During an examination of constraint re-ranking in phonological change, Gess (2003)

presents a three level model (primarily based on the methodological approach of Lexical

Phonology), in which he makes the primary distinction between lexical and post-lexical

components. It is the structure of the post-lexical component which is the most relevant

to the model presented here. Gess proposes that this level is split into two levels:

‘register-dependent’ and ‘register-independent’. Importantly, with respect to

phonological change, the register-dependent level is where the new form is developed

due to a novel re-ranking of applicable OT constraints. Gess (2003:76) notes that the

ranking which produces the new form will be a rarity and will be restricted to a minority

of speech styles, a minority of speakers and, maybe, only feature in frequently occurring

words. The ranking which produces these novel forms will then become more frequent

in a speech community and extend its domain of application with respect to speech styles

and lexical items. Finally, Gess suggests that “the spread of change through speech

contexts and speakers is due to sociolinguistic factors [while] the spread of change

through the lexicon (lexical diffusion) is due to both social factors and frequency”

(2003:77). As a result, this proposal does not deny the role of sociolinguistic factors

300

during language change, particularly the factor of register (even though this is the factor

which both Preston and Bell’s models, outlined above, suggest has the least impact on

variation). The new linguistic forms may be generated in the Realisation and thus, the

notion of register variation would be incorporated in the Sociolinguistics level under the

category of Speech Type. This is in line with Gess’ idea that the new variant is produced

by a novel re-ranking of applicable constraints.

However, this still implies that new variants are generated internally by the linguistic

system but maintained, at least in the early stages, by the sociolinguistic system.

Therefore, one final aspect of change to account for is the possibility of change which is

prompted from outside the linguistic system, such as that generated by contact. In these

cases, the new surface variants may be the result of, for example, long-term

accommodation and therefore, the original influence will be external as opposed to

internal. Cases such as these can be registered directly with the Sociolinguistic level.

When a speaker is introduced to a new form by perceiving its production by another, this

input is then analysed by the phonology and stored among other appropriate exemplars.

The sociolinguistic information will be catalogued (albeit it in less specific terms) along

with the linguistic form and all this information will be available for retrieval when it is

required by the system. Thus, the perception and subsequent perceptual processing of a

new variant creates an exemplar trace in the Sociolinguistics level which will contain

301

information not only relating to sociolinguistic factors, but also the internal processing

information needed in order to actually produce the sound, or an approximation of it.

The more a speaker is exposed to the new form, the more likely they will select it from

the Sociolinguistic level’s inventory as an appropriate form due to motivations such as

stylistic projections (unless divergence from the new forms is desired by the speaker as

opposed to convergence). This increased frequency of usage may prompt reanalysis

within the phonetic stage of the Realisation (either as specific lexical constraints or more

general contextual constraints), perhaps representing the most frequent or most natural

phonological contexts. Once these are established, increased pressure from the

Realisation constraints may cause a shift in Inventory forms either within or across

generations. Conversely, if naturalistic patterns of variation are established through

constraints (such as those associated with Canadian Raising), the Inventory may be left

unaltered and change from the old form to the new one may never take place. Instead,

allophonic variations may become established and entrenched in the language or dialect.

302

As a result, we can propose three types of change:

1. Change which is introduced and maintained by the linguistic system. These

changes will reflect natural processes and motivations.

2. Change which is introduced by the linguistic system, but maintained by the

sociolinguistic system. These changes are the result of novel constraint re-

rankings, which then become established.

3. Change which is introduced externally to the Sociolinguistic level and is

maintained by the sociolinguistic system. These new forms may never become

adopted by the linguistic system in the form of Realisation constraints if, for

example, the contact situation which introduced the innovations diminishes.

Changes of the third type may be the least likely to be adopted by the system since it is

these which need constant exposure, reinforcement and selection in order for contextual

constraints to be established. Therefore, unless they do become established, they

represent only short-term accommodations as opposed to language change through long

term accommodation and contact. In a similar fashion, changes introduced to the system

through novel re-rankings still need to be reinforced by their selection in specific

sociolinguistic contexts. If their application in certain registers, contexts or styles fails to

become established, there is no impetus from the Sociolinguistic level with which to

pressure the linguistic system into maintaining or permanently adopting the ranking.

Finally, change which is motivated and maintained by the linguistic system can be said to

reflect natural motivations within the system as a whole. It is these changes which are

most likely to be stabilised (both linguistically and sociolinguistically), and result in

underlying change, unless they are repeatedly overridden by external sociolinguistic

factors, such as those relating to standardisation and ‘correctness’.

303

This Chapter has discussed how we might represent the interface between phonology and

sociolinguistics. The three-tiered model discussed and presented throughout illustrates

how underlying forms may be selected as part of an optimal underlying system (the

Inventory) through the use of dispersion constraints. These underlying forms are selected

regardless of linguistic context and are used as the input to the second, Realisation, stage.

It is this stage where contextual phonetic information becomes available and allophonic

variation can be generated. This stage has also been shown to accommodate lexically

conditioned as well as articulatory motivated constraints. The final stage is where

sociolinguistic factors, which influence surface variation, are represented (factors such as

the audience and the context of the speech event). Both Inventory and Realisation forms

have trace representations in this level and this is where their sociolinguistic associations

are activated. These units are then evaluated along with any other externally sourced

variants according to sociolinguistic factors.

Finally, the process of language change has been discussed in relation to how types of

change may be reflected in the structure of the three-tiered model. It has been suggested

that three types of change are possible. These changes are sourced either internally or

externally, each with the need for constant reinforcement from internal generation of

constraints and sociolinguistic selection in order to result in change to the underlying

structure.

304

Concluding Remarks

305

CONCLUDING REMARKS

The investigation of Mersea Island English (MIE) has resulted in the discovery of a

number of social and linguistic factors which have influenced the shape and type of

dialect variation present in the data.

We have seen that, as the demographic structure of the Island has developed, there has

been a widening of Islanders’ social network structures. This has resulted in a greater

dependence on the mainland with respect to factors such as education and employment,

as well as more leisurely pursuits, such as shopping and social interaction.

The analysis of life-long Island residents across three generations has demonstrated that it

is not just the socio-demographic characteristics of the Island which have been

susceptible to change, but also the internal structure of the MIE dialect. The diphthongs

//, // and // have been shown to have shifted from their more traditional realisations

of [], [] and [], respectively, which contradict the predictions and generalisations of

both the Diphthong Shift and the Southern Shift. However, these changes are not without

internal complexities.

Canadian Raising type patterns were found for both // and // diphthongs. These

patterns, as discussed in Chapter 9, may be a by-product and highlight the progression of

change from [] to [] and [] to [], as naturalness can account for the initial

preference of lowering the diphthong nuclei before voiced consonants as opposed to

voiceless consonants. Other lowering influences, which help to promote this change

306

were found in the domain of preceding dorsal consonants, and this too was attributed to

contextual influences.

Other patterns highlighted within the overall variation were the preservation of traditional

rounded [] in BOY tokens and the monophthongal variant [] in the lexical item Island.

These were shown, also in Chapter 9, to contrast with respect to performance phrases and

overt dialect comment. Thus, while rounded [] featured in performance phrases and

was open to comments about the characteristics of MIE, Island monophthongisation was

not. This, it was suggested, together with the behaviour of Canadian Raising, helps to

demonstrate different types of social and dialect enregisterment – for example, the

difference between enregistering individual phonological rules or processes and the

realisations of individual lexical items.

The pattern and direction of dialect change and the manipulation of enregistered variants

suggests that phonological conditions and sociolinguistic factors are inter-twined. The

underlying structure of the diphthong system was discussed in Chapter 10 and it was

shown that, when change to the system occurs, dispersion constraints can help illustrate

how perceptual and articulatory distances strive to be maintained. With the integrity of

the underlying system maintained by the phonology (first tier – Inventory), a three-tiered

model of language was presented, which represents the interaction of phonology,

phonetics and sociolinguistics. The second tier (Realisation) allows for phonetic

information to become available for any contextual alternations of underlying forms to

take place. It is here that Canadian Raising, BOY and Island variation can be accounted

307

for and represented. However, neither the Inventory nor Realisation levels allow for

variation influenced by social factors. Therefore, the innovative Sociolinguistic level

retains a trace from the previous levels’ output, as well as any variants in the ambient

speech environment. All these candidates can then be evaluated and selected according

to relevant social factors. The resulting model, thus, equally represents Internal and

External influences on variation, while limiting intra-speaker variation to that only within

the processing capabilities of the phonological system.

The construction of this model also allows for different types of change to be sourced to a

particular level of origin. For example, externally-driven changes (such as that from

long-term accommodation) can be shown to originate in the Sociolinguistic level before it

filters upwards through the system until the underlying structure is prompted to re-

evaluate and adjust itself accordingly. Conversely, changes which may be prompted by

natural linguistic motivations (such as coarticulation or changes towards the unmarked)

can be shown to originate in the Inventory or Realisation levels. If the innovative forms

are selected and used by the Sociolinguistic level, they will become reinforced and

adopted by the system as linguistic change or allophonic variation.

As a result, this model is able to represent phonological change as well as linguistic

variation (both allophonic and sociolinguistic). The implications of such a model allow

for the future re-examination and representation of existing data, as well as the

development of innovative methodologies which can allow for the socio-phonological

interface to be investigated further.

308

Bibliography

309

(2000) The New Penguin Dictionary; Penguin Books

Agha, A. (2003) The Social Life of Cultural Values; Language and Communication 23:

231-273

Altendorf, U. (2003) Estuary English: levelling at the interface of RP and South-Eastern

British English; Gunter Narr Verlag (Germany)

Anttila, A. (2002) Variation and Phonological Theory; in Chambers, J., Trudgill, P. and

Schilling-Estes, N. (eds) The Handbook of Language Variation and Change; Blackwell

(UK)

Anttila, A. (2007) Variation and Optionality; in de Lacey, P. (ed) The Cambridge

Handbook of Phonology; Cambridge University Press (UK)

Beal, J. (2009) Enregisterment, Commodification and Historical Context: “Geordie”

versus “Sheffield”; American Speech 84: 138-156

Bell, A. (1984) Language Style as Audience Design; Language in Society 13: 145-204

Bell, A. (2001) Back in Style: reworking audience design; in Eckert, P. and Rickford, J.

(eds) Style and Sociolinguistic Variation; Cambridge University Press (UK)

Blake, R. and Josey, M. (2003) The /ay/ Diphthong in a Martha’s Vineyard Community:

what can we say 40 years after Labov?; Language and Society 32: 451-485

Boersma, P. and Hayes, B. (1999) Empirical Tests of the Gradual Learning Algorithm;

Rutgers Optimality Archive, ROA 348-1099 (http://roa.rutgers.edu)

Boersma, P. and Hayes, B. (2001) Empirical Tests of the Gradual Learning Device;

Linguistic Inquiry 32: 45-86

Britain, D. (1991) Dialect and Space: a geolinguistic study of speech variables in the

Fens; Unpublished PhD, University of Essex (UK)

Britain, D. (1995) The Sociolinguistic Development of Canadian Raising in the Fens;

Essex Research Reports in Linguistics 6: 1-53, University of Essex (UK)

Britain, D. (1997) Dialect Contact and Phonological Reallocation: ‘Canadian Raising’

in the English Fens; Language in Society 26: 15-46

Britain, D. (2002a) Space and Spatial Diffusion; in Chambers, J., Trudgill, P. and


(UK)

http://roa.rutgers.edu/

310

Britain, D. (2002b) The British History of New Zealand English; Essex Research Reports

in Linguistics 41: 1-41 (Reprinted as 'Where did New Zealand English come from ?' in

Bell, A., Harlow, R. and Starks, D (eds) (2005) The Languages of New Zealand; Victoria

University Press (NZ))

Britain, D. (2002c) Phoenix from the Ashes: the death, contact and birth of dialects in

England; Essex Research Reports in Linguistics 41: 42-72 (Reprinted as 'One foot in the

grave?: Dialect death, dialect contact and dialect birth in England' (2010) International

Journal of Sociology of Language 196/197: 121-155)

Britain, D. (2008) On the Wrong Track? A non-standard history of non-standard /au/ in

English; Essex Research Reports in Linguistics 57: 79-114

Britain, D. (2010) Supralocal Regional Dialect Levelling; in Llamas, C. and Watt, D.

(eds) Language and Identities; Edinburgh University Press (UK)

Britain, D. (forthcoming) Diffusion; in Bergs, A. and Brinton, L. (eds) Historical

Linguistics of English; Mouton de Gruyter (Germany)

Britain, D. and Sudbury, A. (2010) South Atlantic Ocean: Falkland Island English; in

Schreier, D., Trudgill, P., Schneider, E and Williams, J. (eds) Lesser-Known Varieties of

English; Cambridge University Press (UK)

Bybee, J. (2002) Word Frequency and Context of Use in the Lexical Diffusion of

Phonetically Conditioned Sound Change; Language Variation and Change 14: 261-290

Bybee, J. (2006) From Usage to Grammar: the mind’s response to repetition; Language

82: 711-733

Chambers, J. (1989) Canadian Raising: blocking, fronting etc; American Speech 64:75-

88

Chambers, J. (2002) Studying Language Variation: an informal epistemology; in

Chambers, J., Trudgill, P. and Schilling-Estes, N. (eds) The Handbook of Language

Variation and Change; Blackwell (UK)

Cheshire, J. and Fox, S. (2007) Innovation in the Quotative System of London

Adolescents; Paper presented at NWAV 36, University of Pennsylvania (USA) -

(Available at:

http://www.lancs.ac.uk/fss/projects/linguistics/innovators/documents/NWAV_36_Fox_C

heshire_000.pdf)

http://www.lancs.ac.uk/fss/projects/linguistics/innovators/documents/NWAV_36_Fox_Cheshire_000.pdf

http://www.lancs.ac.uk/fss/projects/linguistics/innovators/documents/NWAV_36_Fox_Cheshire_000.pdf

311

Cheshire, J., Gillett, A., Kerswill, P. and Williams, A. (1999) The Role of Adolescents in

Dialect Levelling; Final report submitted to the ESRC, June 1999 (ESRC Ref.

R000236180)

Chirrey, D. (1999) Edinburgh: descriptive material; in Foulkes, P. and Docherty, G.

Urban Voices; Arnold (UK)

Chomsky, N. and Halle, M. (1968) The Sound Pattern of English; Harper and Row

(USA)

Clarke, J., Yallop, C. and Fletcher, J. (2007) An Introduction to Phonetics and

Phonology; Blackwell (UK)

Coates, J. (2004) Women, Men and Language – 3rd

Edition; Pearson Education Ltd (UK)

Collins, B. and Mees, I. (1990) The Phonetics of Cardiff English; in Coupland, N. (ed)

English in Wales; Multilingual Matters (UK)

Cruttenden, A. (2001) Gimson’s Pronunciation of English: 6th

Edition; Arnold (UK)

D’Arcy, A. (2007) Like and Language Ideology: disentangling fact from fiction;

American Speech 82: 386-419

Docherty, G. and Foulkes, P. (1999) Derby and Newcastle: instrumental phonetics and

variationist studies; in Foulkes, P. and Docherty, G. Urban Voices; Arnold (UK)

Docherty, G., Foulkes, P., Milroy, J., Milroy, L. and Walshaw, D. (1997) Descriptive

Adequacy in Phonology: a variationist perspective; Journal of Linguistics 33: 275-310

Drager, K. (2009) A Sociophonetic Ethnography of Selwyn Girls’ High; Unpublished

PhD, University of Canterbury (NZ)

Eckert, P. (1989) Jocks and Burnouts: social categories and identity in the high school;

Teachers College Press (USA)

Eckert, P. (1997) Age as a Sociolinguistic Variable; in Coulmas, F. (ed) The Handbook of

Sociolinguistics; Blackwell (UK)

Eckert, P. (1998) Gender as a Sociolinguistic Variable; in Coates, J. (ed) Language and

Gender: a reader; Blackwell (UK)

Eckert, P. (2000) Linguistic Variation as Social Practice: the linguistic construction of

identity in Belton High; Blackwell (UK)

312

Ellis, A. (1889) On Early English Pronunciation, Part V – the existing phonology of

English Dialects compared to that of West Saxon speech; English Text Society [unaltered

re-print 1999] Woodbridge (UK)

Eustace, S. (1969) The Meaning of the Palaeotype in A J Ellis’ ‘On Early English

Pronunciation’; Transactions of the Philological Society 67: 31-79

Flemming, E. (2004) Contrast and Perceptual Distinctiveness; in Hayes, B., Kirchner, R.

and Steriade, D. (eds) Phonetically Based Phonology; Cambridge University Press (UK)

Flemming, E. (2006) The Role of Distinctiveness Constraints in Phonology;

(http://web.mit.edu/flemming/www/paper/inventories.pdf last accessed September 2011)

Foulkes, P. and Docherty, G. (eds) (1999) Urban Voices; Arnold (UK)

Fox, S. (2000) Estuary English’: Fact or Fiction?; Paper presented at SociolinguistEssex

V, University of Essex (UK)

Fox, S. (2007) The Demise of Cockneys? Language Change in London's 'Traditional'

East End; Unpublished PhD, University of Essex (UK)

Gess, R. (2003) On Re-Ranking and Explanatory Adequacy and a Constraint-Based

Theory of Phonological Change; in Holt, D. (ed) Optimality Theory and Language

Change; Kluwer Academic Publishers (Netherlands)

Hammond, M. (1994) An OT Account of Variability in Walmatjari Stress; Rutgers

Optimality Archive, ROA-20-0794 (http://roa.rutgers.edu)

Hammond, M. (1999) The Phonology of English; a prosodic optimality-theoretic

approach; Oxford University Press (UK)

Hymes, D. (1972) On Communicative Competence; in Pride, J. and Holmes, J. (eds)

Sociolinguistics: selected readings; Penguin (UK)

Hymes, D. (2001) On Communicative Competence; Duranti, A. (ed) Linguistic

Anthropology; a reader; Blackwell (UK)

Jay, B. (1996) Business as Usual; Mersea Island Museum Publications (UK)

Jay, B. (2005) Isle of Mersea; B.M. Jay, Personal Publication

Johnson, W. and Britain, D. (2007) L-vocalisation as a natural phenomenon:

Explorations in Sociophonology; Language Sciences 29: 294-315

Johnstone, B., Andrus, J. and Danielson, A. (2006) Mobility, Indexicality and the

Enregisterment of “Pittsburghese”; Journal of English Linguistics 34: 77-104

http://web.mit.edu/flemming/www/paper/inventories.pdf


313

Karbacz, E. (1999) A Short History of Mersea; Mersea Island Museum Publications (UK)

Kerswill, P. (2004) Koineization and Accommodation; in Chambers, J., Trudgill, P. and


(UK)

Kerswill, P., Torgersen, E. and Fox, S. (2006) Innovation in Inner-London Teenage

Speech; Paper presented at NWAV 35 (USA)

Kerswill, P. and Williams, A. (2000) Creating a New Town Koine: children and

language change in Milton Keynes; Language and Society 29: 65-115

Kiparsky, P. (1968) Linguistic Universals and Linguistic Change; in Bach, E. and Harms,

R. (eds) Universals in Linguistic Theory; Holt, Rinehart and Winston (USA)

Kiparsky, P. (1993) An OT Perspective on Phonological Variation; Rutgers Optimality

Workshop 1 (1993), (http://www.stanford.edu/~kiparsky/Papers/nwave94.pdf last

accessed September 2011)

Kiparsky, P. (2000) Opacity and Cyclicity; The Linguistic Review 17:315-376

Kiparsky, (2008) Fenno-Swedish Quantity: contrast in Stratal OT; in Nevins, A. and

Vaux, B. (ed) Rules, Constraints and Phonological Phenomena; Oxford University Press

(UK)

Kirchner, R. (2001) Phonological Contrast and Articulatory Effort; in Lombardi, L. (ed)

Segmental Phonology in Optimality Theory; Cambridge University Press (UK)

Kirchner, R. (2004) Consonant Lenition; in Hayes, B., Kirchner, R. and Steriade, D. (eds)

Phonetically Based Phonology; Cambridge University Press (UK)

Kokeritz, H. (1932) The Phonology of the Suffolk Dialect, descriptive and historical;

Inaugural Dissertation, Philosophical Faculty of the University of Uppsala (Sweden)

Kurath, H. and Lowman, G. (1970) The Dialect Structure of Southern England;

University of Alabama Press (USA)

Labov, W. (1972) Sociolinguistic Patterns; Blackwell (UK)

Labov, W. (1994) Principles of Linguistic Change: Internal Factors; Blackwell (UK)

Ladefoged, P. (2006) A Course in Phonetics: 5th

Edition; Thomson Wadsworth (USA)

Lass, R. (1987) The Shape of English; J.M. Dent and Sons Ltd (UK)

http://www.stanford.edu/~kiparsky/Papers/nwave94.pdf

314

Laver, J. (1994) Principles of Phonetics; Cambridge University Press (UK)

Macaulay, R. (2001) You’re like why not: the quotative expressions in Glasgow

adolescents; Journal of Sociolinguistics 5: 3-21

Mathisen, A. (1999) Sandwell, West Midlands: ambiguous perspectives on gender

patterns and models of change; in Foulkes, P. and Docherty, G. Urban Voices; Arnold

(UK)

McCarthy, J. (2002) A Thematic Guide to Optimality Theory; Cambridge University

Press (UK)

McCarthy, J. (2006) GEN, EVAL and Phonological Opacity;

http://people.umass.edu/jjmccart/phonologyfest/mccarthy-phonologyfest.pdf - last

accessed September 2011

McCarthy, J. (2007) Hidden Generalisations: phonological opacity in optimality theory;

Equinox (UK)

McCarthy, J. (2008) Doing Optimality Theory: applying theory to data; Blackwell (UK)

McCarthy, J. and Prince, A. (1995) Faithfulness and Reduplicative Identity; Rutgers

Optimality Archive, ROA-60 (http://roa.rutgers.edu)

McMahon, A. (2007) Who’s Afraid of the Vowel Shift Rule?; Linguistic Science 29: 341-

359

Mees, I. and Collins, B. (1999) Cardiff: a real-time study of glottalisation; in Foulkes, P.

and Docherty, G. (eds) Urban Voices; Arnold (UK)

Mendoza-Denton, N. (2008) Homegirls; Blackwell (UK)

Meyerhoff, M. (2002) Communities of Practice; in Chambers, J., Trudgill, P. and


(UK)

Milroy, J. (1992) Linguistic Variation and Change: on the historical sociolinguistics of

English; Blackwell (UK)

Milroy, J., Milroy, L., Hartley, S. and Walshaw, D. (1994) Glottal Stops and Tyneside

Glottalisation: competing patterns of variation and change in British English; Language

Variation and Change 6: 327-357

Milroy, L. (1987) Language and Social Networks; 2nd

Edition; Blackwell (UK)

http://people.umass.edu/jjmccart/phonologyfest/mccarthy-phonologyfest.pdf


315

Milroy, L. and Gordon, M. (2003) Sociolinguistics: methods and interpretations;

Blackwell (UK)

Minkova, D. and Stockwell, R. (2003) English Vowel Shifts and ‘Optimal’ Diphthongs:

is there a logical link?; in Holt, D. (ed) Optimality Theory and Language Change; Kluwer

Academic Publishers (Netherlands)

Ogura, M. (1987) Historical English Phonology; Kenkysha (Tokyo)

Orton, H. (1962) Survey of English Dialects – Introduction; Arnold (UK)

Orton, H. and Tilling, P. (1969) Survey of English Dialects, Volume 3 – The East

Midland Counties and East Anglia, Part 1; Arnold (UK)





Padgett, J. (1995) Partial Class Behavior and Nasal Place Assimilation; Proceedings of

the Arizona Phonology Conference: Workshop on Features in Optimality Theory, Coyote

Working Papers, University of Arizona (USA)

Pierrehumbert, J. (2001) Exemplar Dynamics: word frequency, lenition and contrast; in

Bybee, J. and Hopper, P. (eds) Frequency and the Emergence of Linguistic Structure;

John Benjamins (USA)

Pope, J., Meyerhoff, M. and Ladd, R. (2007) Forty Years of Change on Martha’s

Vineyard; Journal of Linguistic Society of America 83: 615-627

Preston, D. (1991) Sorting out the Variables in Sociolinguistic Theory; American Speech

66: 33-56

Prince, A. and Smolensky, P. (1993) Optimality Theory: constraint interaction in

generative grammar; Rutgers Optimality Archive, ROA-537 (http://roa.rutgers.edu)

Prince, A. and Smolensky, P. (2004) Optimality Theory: constraint interaction in

generative grammar; Blackwell (UK)


316

Przedlacka, J. (2001) Estuary English and RP: some recent findings; Studia Anglica

Posnaniensia 36: 35-50

Ryfa, J. (2005) Chavs and Grungers: the creation of distinct speech styles by two hostile

communities of practice in Colchester, Essex; Unpublished MA Dissertation, University

of Essex (UK)

Ryfa, J. (forthcoming) Englishes of London and the Southeast of England; in Hopkins, T.

and McKenny, J. (eds) World Englishes: Volume 1; Continuum International (UK)

Schilling-Estes, N. (1998) Investigating “self-conscious” speech: the performance

register in Ocracoke English; Language in Society 27: 53-83

Schilling-Estes, N. (2002) On the Nature of Isolated and Post-Isolated Dialects:

innovation, variation and differentiation; Journal of Sociolinguistics 6: 64-85

Schilling-Estes, N. and Wolfram, W. (1999) Alternative Models of Dialect Death:

dissipation vs. concentration; Language 75: 486-521

Schreier, D. (2008) St Helenian English: origins, evolution and variation; John

Benjamins (USA)

Schreier, D. (2010a) St Helenian English; in Schreier, D., Trudgill, P., Schneider, E. and

Williams, J. (eds) The Lesser-Known Varieties of English; Cambridge University Press

(UK)

Schreier, D. (2010b) Tristan da Cunha English; in Schreier, D., Trudgill, P., Schneider, E.

and Williams, J. (eds) The Lesser-Known Varieties of English; Cambridge University

Press (UK)

Searle, J. (1979) Expression and Meaning; Cambridge University Press (UK)

Searle, J. (1999) Mind, Language and Society; Weidenfeld and Nicolson (UK)

Sivertsen, E. (1960) Cockney Phonology; Humanities Press (USA)

317

Stockwell, R. (1975) Problems in the Interpretation of the Great English Vowel Shift; in

Goyvaerts, D. and Pullum, G. Essays on the Sound Pattern of English; E. Story-Scientia

(Belgium)

Stoddard, J., Upton, C. and Widdowson, J. (1999) Sheffield in the 1990s:revisiting the

concept of NORMs; in Foulkes, P. and Docherty, G. (eds) Urban Voices; Arnold (UK)

Strang, B. (1970) A History of English; Methuen and Co. Ltd (UK)

Stuart-Smith, J. (1999) Glasgow: accent and voice quality; in Foulkes, P. and Docherty,

G. (eds) Urban Voices; Arnold (UK)

Sudbury, A. (2000) Dialect Contact and Koineisation in the Falkland Islands;

Unpublished PhD, University of Essex (UK)

Sudbury, A. and Britain, D. (2000) Is Falkland Island English Linguistically a Southern

Hemisphere Variety?; Essex Research Reports in Linguistics 28: 33-59 (Reprinted as

Sudbury, A. (2001) ‘Is Falkland Island English a southern hemisphere variety?’; English

Worldwide 22: 55-80).

Tagliamonte, S. and Hudson, R. (1999) Be like et al. beyond America: the quotative

system in British and Canadian Youth; Journal of Sociolinguistics 3:147-172

Tatham, M. and Morton, K. (2006) Speech Production and Perception; Palgrave

MacMillan (UK)

Tollfree, L. (1999) South East London English: discrete versus continuous modelling of

consonantal reduction; in Foulkes, P. and Docherty, G. (eds) Urban Voices; Arnold (UK)

Torgersen, E. and Kerswill, P. (2004) Internal and External Motivation in Phonetic

Change: dialect levelling outcomes for an English Vowel shift; Journal of

Sociolinguistics 8: 23-53

Trudgill, P. (1974) The Social Differentiation of English in Norwich; Cambridge

University Press (UK)

Trudgill, P. (1986) Dialects and Contact; Blackwell (UK)

Trudgill, P. (1988) Norwich Revisited: recent linguistic changes in an English urban

dialect; English World-Wide 9: 33-49

Trudgill, P. (1999) Norwich: endogenous and exogenous linguistic change; in Foulkes, P.

and Docherty, G. (eds) Urban Voices; Arnold (UK)

Wakelin, M. (1977) English Accents: an introduction (revised addition); Athlone Press

(UK)

http://openlibrary.org/publishers/E._Story-Scientia

318

Weinreich, U., Labov, W. and Herzog, M. (1968) Empirical Foundations for a Theory of

Language Change; in Lehman, W. P. and Malkiel, Y. (eds) Directions for Historical

Linguistics; University of Texas Press (USA)

Wells, J. (1982a) Accents of English: an introduction; Cambridge University Press (UK)

Wells, J. (1982b) Accents of English: the British Isles; Cambridge University Press (UK)

Wells, J. (1982c) Accents of English: Beyond the British Isles; Cambridge University

Press (UK)

Williams, A. and Kerswill, P. (1999) Dialect Levelling: change and continuity in Milton

Keynes, Reading and Hull; in Foulkes, P. and Docherty, G. (eds) Urban Voices; Arnold

(UK)

Wolfram, W. (2008) When Islands lose Dialects; the case of the Ocracoke Brogue;

Shima: The International Journal of Research into Island Cultures 2: 1-13

Wolfram, W. and Schilling-Estes, N. (2006) American English: 2nd

Edition; Blackwell

(UK)

Wright, J. (1968) The English Dialect Grammar; Oxford University Press (UK)

Yamada, N. (1984) On Characterising the English Great Vowel Shift; Lingua 62: 43-69

319

Appendix A

320

Appendix A

In this appendix, you will find the complete numerical data sets pertaining to the analysis

of the variable () (see Chapter 4).

A.3a

Number of () tokens according to age and gender

Total

Young

Male 1 4 257 105 367

Female 2 288 73 363

Old

Male 18 8 579 217 200 14 1036

Female 1 283 195 346 35 860

Museum

Male 16 16 241 27 15 1 316

A.3b


Young

Male 0.27 1.09 70.03 28.61

Female 0.55 79.34 20.11

Old

Male 1.74 0.77 55.89 20.95 19.31 1.35

Female 0.12 32.91 22.67 40.23 4.07

Museum

Male 5.06 5.06 76.27 8.54 4.75 0.32

321

A.4a

Overall number of () tokens according to syllable type

Total

Open 1 4 233 24 125 9 233 65 55 749

Closed 13 16 25 668 179 265 46 506 302 173 2193

A.4b

Overall percentage of () tokens according to syllable type

Open 0.13 0.53 31.11 3.20 16.69 1.20 31.11 8.68 7.34

Closed 0.59 0.73 1.14 30.46 8.16 12.08 2.10 23.07 13.77 7.89

A.5a

Overall number of () tokens according to offglide duration and syllable type

Full Short aa Total

Open 596 98 55 749

Closed 1468 552 173 2193

Total 2064 650 228 2942

A.5b

Overall percentage of () variants according to offglide duration and syllable type

Full Short aa

Open 79.57 13.08 7.34

Closed 66.94 25.17 7.89

322

A.6a

Number of () tokens according to age, syllable type and offglide duration

Full Short aa Total

Young

Open 124 39 43 206

Closed 243 146 135 524

Old

Open 415 56 12 483

Closed 1035 341 37 1413

Museum

Open 57 3 0 60

Closed 190 65 1 256

A.6b

Percentage of () variants according to age, syllable type and offglide duration

Full Short aa

Young

Open 60.19 18.93 20.87

Closed 46.37 27.86 25.76

Old

Open 85.92 11.59 2.48

Closed 73.25 24.13 2.62

Museum

Open 95.00 5.00 0.00

Closed 74.22 25.39 0.39

323

A.7a

Overall number of () tokens according to preceding manner of articulation

Total

Stop 29 19 461 153 308 79 1049

Fricative 1 173 99 274 34 581

Nasal 5 144 88 173 33 443

Approximant 58 29 159 26 272

/sk/ 2 2 4

Vowel 1 1

None 5 265 76 190 56 592

A.7b

Overall percentage of () variants according to preceding manner of articulation

Stop 2.76 1.81 43.95 14.59 29.36 7.53

Fricative 0.17 29.78 17.04 47.16 5.85

Nasal 1.13 32.51 19.86 39.05 7.45

Approximant 21.32 10.66 58.46 9.56

/sk/ 50.00 50.00

Vowel 100

None 0.84 44.76 12.84 32.09 9.46

A.8a

Overall number of () tokens according to preceding place of articulation

Total

Labial 1 15 246 66 186 59 573

Coronal 31 4 441 207 476 93 1252

Dorsal 2 14 11 39 3 69

/h/ 1 135 85 213 17 451

None 5 265 76 190 56 592

A.8b

Overall percentage of () variants according to preceding place of articulation

Labial 0.17 2.62 42.93 11.52 32.46 10.30

Coronal 2.48 0.32 35.22 16.53 38.02 7.43

Dorsal 2.90 20.29 15.94 56.52 4.35

/h/ 0.22 29.93 18.85 47.23 3.77

None 0.84 44.76 12.84 32.09 9.46

324

A.9a

Overall number of () tokens according to following manner of articulation

Total

Stop 1 16 442 116 305 101 981

Fricative 1 112 55 151 11 330

Nasal 28 4 287 149 396 58 922

Approximant 3 3

Affricate 1 7 1 9

Schwa 9 4 7 1 21

Flap 4 35 9 4 10 62

None 5 218 112 233 46 614

A.9b

Overall percentage of () variants according to following manner of articulation

Stop 0.10 1.63 45.06 11.82 31.09 10.30

Fricative 0.30 33.94 16.67 45.76 3.33

Nasal 3.04 0.43 31.13 16.16 42.95 6.29

Approximant 100.00

Affricate 11.11 77.78 11.11

Schwa 42.86 19.05 33.33 4.76

Flap 6.45 56.45 14.52 6.45 16.13

None 0.81 35.50 18.24 37.95 7.49

A.10a

Overall number of () tokens according to following place of articulation

Total

Labial 2 1 7 10

Coronal 28 10 491 249 645 99 1522

Glottal 1 15 384 79 214 82 775

Schwa 9 4 7 1 21

None 5 218 112 233 46 614

A.10b

Overall percentage of () variants according to following place of articulation

Labial 20.00 10.00 70.00

Coronal 1.84 0.66 32.26 16.36 42.38 6.50

Glottal 0.13 1.94 49.55 10.19 27.61 10.58

Schwa 42.86 19.05 33.33 4.76

None 0.81 35.50 18.24 37.95 7.49

325

A.11a

Overall number of () tokens according to following voice

Total

Voiced 28 8 371 176 457 77 1117

Voiceless 1 17 515 157 416 105 1211

None 5 0 218 112 233 46 614

A.11b

Overall percentage of () variants according to following voice

Voiced 2.51 0.72 33.21 15.76 40.91 6.89

Voiceless 0.08 1.40 42.53 12.96 34.35 8.67

None 0.81 0.00 35.50 18.24 37.95 7.49

A.12a

Number of () tokens according to following voice by Museum speakers

Total

Voiced 14 6 77 15 12 1 125

Voiceless 0 10 119 6 2 0 137

None 2 0 45 6 1 0 54

A.12b

Percentage of () variants according to following voice by Museum speakers

Voiced 11.20 4.80 61.60 12.00 9.60 0.80

Voiceless 0.00 7.30 86.86 4.38 1.46 0.00

None 3.70 0.00 83.33 11.11 1.85 0.00

326

A.13a

Number of () tokens according to following voice by Older speakers

Total

Voiced 14 2 306 144 270 20 756

Voiceless 1 7 396 149 184 19 756

None 3 0 173 105 93 10 384

A.13b

Percentage of () variants according to following voice by Older speakers

Voiced 1.85 0.26 40.48 19.05 35.71 2.65

Voiceless 0.13 0.93 52.38 19.71 24.34 2.51

None 0.78 0.00 45.05 27.34 24.22 2.60

A.14a

Number of () tokens according to following voice by Younger speakers

Total

Voiced 0 0 1 3 176 56 236

Voiceless 0 0 0 2 230 86 318

None 0 0 0 1 139 36 176

A.14b

Percentage of () variants according to following voice by Younger speakers

Voiced 0.00 0.00 0.42 1.27 74.58 23.73

Voiceless 0.00 0.00 0.00 0.63 72.33 27.04

None 0.00 0.00 0.00 0.57 78.98 20.45

327

From preceding Phonological Environment Chapter 4.3.2

Museum

A.15a

Number of () tokens by Museum speakers according to preceding manner of

articulation

Total

Stop 14 12 99 8 7 1 141

Fricative 1 30 4 1 36

Nasal 2 29 4 1 36

Approximant 15 6 4 25

/sk/ 0

Vowel 0

None 3 68 5 2 78

A.15b

Percentage of () variants by Museum speakers according to preceding manner of

articulation

Stop 9.93 8.51 70.21 5.67 4.96 0.71

Fricative 2.78 83.33 11.11 2.78

Nasal 5.56 80.56 11.11 2.78

Approximant 60.00 24.00 16.00

/sk/

Vowel

None 3.85 87.18 6.41 2.56

328

Old

A.16a

Number of () tokens by Older speakers according to preceding manner of articulation

Total

Stop 15 7 361 141 166 24 714

Fricative 143 93 106 6 348

Nasal 3 115 84 89 7 298

Approximant 43 23 97 5 168

/sk/ 2 2 4

Vowel 1 1

None 2 197 71 86 7 363

A.16.b

Percentage of () variants by Older speakers according to preceding manner of

articulation

Stop 2.10 0.98 50.56 19.75 23.25 3.36

Fricative 41.09 26.72 30.46 1.72

Nasal 1.01 38.59 28.19 29.87 2.35

Approximant 25.60 13.69 57.74 2.98

/sk/ 50.00 50.00

Vowel 100.00

None 0.55 54.27 19.56 23.69 1.93

329

Young

A.17a

Number of () tokens by Younger speakers according to preceding manner of

articulation

Total

Stop 1 4 135 54 194

Fricative 2 167 28 197

Nasal 83 26 109

Approximant 58 21 79

/sk/ 0

Vowel 0

None 102 49 151

A.17b

Percentage of () variants by Younger speakers according to preceding manner of

articulation

Stop 0.52 2.06 69.59 27.84

Fricative 1.02 84.77 14.21

Nasal 76.15 23.85

Approximant 73.42 26.58

/sk/

Vowel

None 67.55 32.45

330

Museum

A.18a

Number of () tokens by Museum speakers according to preceding place of articulation

Total

Labial 9 63 5 3 80

Coronal 14 3 84 12 7 120

Dorsal 2 4 2 3 1 12

/h/ 1 22 3 26

None 3 68 5 2 78

A.18b

Percentage of () variants by Museum speakers according to preceding place of

articulation

Labial 11.25 78.75 6.25 3.75

Coronal 11.67 2.50 70.00 10.00 5.83

Dorsal 16.67 33.33 16.67 25.00 8.33

/h/ 3.85 84.62 11.54

None 3.85 87.18 6.41 2.56

331

Old

A.19a

Number of () tokens by Older speakers according to preceding place of articulation

Total

Labial 1 6 182 58 86 15 348

Coronal 17 1 357 193 261 22 851

Dorsal 10 9 25 1 45

/h/ 113 81 86 4 284

None 2 197 71 86 7 363

A.19b

Percentage of () variants by Older speakers according to preceding place of

articulation

Labial 0.29 1.72 52.30 16.67 24.71 4.31

Coronal 2.00 0.12 41.95 22.68 30.67 2.59

Dorsal 22.22 20.00 55.56 2.22

/h/ 39.79 28.52 30.28 1.41

None 0.55 54.27 19.56 23.69 1.93

332

Young

A.20a

Number of () tokens by Younger speakers according to preceding place of articulation

Total

Labial 1 3 97 44 145

Coronal 2 208 71 281

Dorsal 11 1 12

/h/ 1 127 13 141

None 102 49 151

A.20b

Percentage of () variants by Younger speakers according to preceding place of

articulation

Labial 0.69 2.07 66.90 30.34

Coronal 0.71 74.02 25.27

Dorsal 91.67 8.33

/h/ 0.71 90.07 9.22

None 67.55 32.45

333

From Following Phonological Environment Chapter 4.3.3

Museum

A.21a

Number of () tokens by Museum speakers according to following manner of

articulation

Total

Stop 9 94 5 2 110

Fricative 1 28 2 31

Nasal 14 3 59 12 12 1 101

Approximant 0

Affricate 1 1

Schwa 0

Flap 3 14 2 19

None 2 45 6 1 54

A.21b

Percentage of () variants by Museum speakers according to following manner of

articulation

Stop 8.18 85.45 4.55 1.82

Fricative 3.23 90.32 6.45

Nasal 13.86 2.97 58.42 11.88 11.88 0.99

Approximant

Affricate 100.00

Schwa

Flap 15.79 73.68 10.53

None 3.70 83.33 11.11 1.85

334

Old

A.22a

Number of () tokens by Older speakers according to following manner of articulation

Total

Stop 1 7 348 109 140 20 625

Fricative 97 40 71 2 210

Nasal 14 1 228 134 234 15 626

Approximant 2 2

Affricate 3 3

Schwa 9 4 3 16

Flap 1 20 6 1 2 30

None 3 173 105 93 10 384

A.22b

Percentage of () variants by Older speakers according to following manner of

articulation

Stop 0.16 1.12 55.68 17.44 22.40 3.20

Fricative 46.19 19.05 33.81 0.95

Nasal 2.24 0.16 36.42 21.41 37.38 2.40

Approximant 100.00

Affricate 100.00

Schwa 56.25 25.00 18.75

Flap 3.33 66.67 20.00 3.33 6.67

None 0.78 45.05 27.34 24.22 2.60

335

Young

A.23a

Number of () tokens by Younger speakers according to following manner of

articulation

Total

Stop 2 163 81 246

Fricative 80 9 89

Nasal 2 151 42 195

Approximant 1 1

Affricate 4 1 5

Schwa 4 1 5

Flap 1 1 3 8 13

None 1 139 36 176

A.23b

Percentage of () variants by Younger speakers according to following manner of

articulation

Stop 0.81 66.26 32.93

Fricative 89.89 10.11

Nasal 1.03 77.44 21.54

Approximant 100.00

Affricate 80.00 20.00

Schwa 80.00 20.00

Flap 7.69 7.69 23.08 61.54

None 0.57 78.98 20.45

336

Museum

A.24a

Number of () tokens by Museum speakers according to following place of articulation

Total

Labial 2 2

Coronal 14 7 109 17 13 1 161

Glottal 9 85 4 1 99

Schwa 0

None 2 45 6 1 54

A.24b

Percentage of () variants by Museum speakers according to following place of

articulation

Labial 100.00

Coronal 8.70 4.35 67.70 10.56 8.07 0.62

Glottal 9.09 85.86 4.04 1.01

Schwa

None 3.70 83.33 11.11 1.85

337

Old

A.25a

Number of () tokens by Older speakers according to following place of articulation

Total

Labial 1 1

Coronal 14 3 394 215 375 26 1027

Glottal 1 6 299 74 75 13 468

Schwa 9 4 3 16

None 3 173 105 93 10 384

A.25b

Percentage of () variants by Older speakers according to following place of

articulation

Labial 100.00

Coronal 1.36 0.29 38.36 20.93 36.51 2.53

Glottal 0.21 1.28 63.89 15.81 16.03 2.78

Schwa 56.25 25.00 18.75

None 0.78 45.05 27.34 24.22 2.60

338

Young

A.26a

Number of () tokens by Younger speakers according to following place of articulation

Total

Labial 7 7

Coronal 1 4 257 72 334

Glottal 1 138 69 208

Schwa 4 1 5

None 1 139 36 176

A.26b

Percentage of () variants by Younger speakers according to following place of

articulation

Labial 100.00

Coronal 0.30 1.20 76.95 21.56

Glottal 0.48 66.35 33.17

Schwa 80.00 20.00

None 0.57 78.98 20.45

339

Appendix B

340

Appendix B


of the variable () (see Chapter 6).

341

B.0a


o Total

Young 33 2 115 6 1021 104 11 3 1 9 5 1310

Old 24 18 433 1 322 6 1150 34 13 1 4 2 2008

Museum 35 27 188 23 27 1 1 302

Total 59 45 654 3 460 12 2198 139 24 4 2 13 7 3620

B.0b


o Young 0.0 0.0 2.5 0.2 8.8 0.5 77.9 7.9 0.8 0.2 0.1 0.7 0.4

Old 1.2 0.9 21.6 0.0 16.0 0.3 57.3 1.7 0.6 0.0 0.0 0.2 0.1

Museum 11.6 8.9 62.3 0.0 7.6 0.0 8.9 0.3 0.0 0.0 0.3 0.0 0.0

342

B.1

Number of () tokens according to diphthong type and age

Total

Young 0 0 35 121 1125 11 4 14 1310

Old 24 18 434 328 1184 13 1 6 2008

Museum 35 27 188 23 28 0 1 0 302

Total 59 45 657 472 2337 24 6 20 3620

B.2a


Total

Young

Male 0 0 19 91 505 9 2 14 640

Female 0 0 16 30 620 2 2 0 670

Old

Male 21 11 371 185 435 0 0 5 1028

Female 3 7 63 143 749 13 1 1 980

Museum

Male 35 27 188 23 28 0 1 0 302

B.2b


Young

Male 0.0 0.0 3.0 14.2 78.9 1.4 0.3 2.2

Female 0.0 0.0 2.4 4.5 92.5 0.3 0.3 0.0

Old

Male 2.0 1.1 36.1 18.0 42.3 0.0 0.0 0.5

Female 0.3 0.7 6.4 14.6 76.4 1.3 0.1 0.1

Museum

Male 11.6 8.9 62.3 7.6 9.3 0.0 0.3 0.0

343

B.3a

Overall number of () tokens according to syllable type

Total

Open 5 15 95 84 549 7 3 0 758

Closed 54 30 562 388 1788 17 3 20 2862

B.3b

Overall percentage of () variants according to syllable type

Open 0.66 1.98 12.53 11.08 72.43 0.92 0.40 0.00

Closed 1.89 1.05 19.64 13.56 62.47 0.59 0.10 0.70

B.4a

Number of () tokens according to syllable type and age

Total

Young (1310, total)

Open 0 0 7 16 282 5 2 0 312

Closed 0 0 28 105 843 6 2 14 998

Old (2008, total)

Open 3 8 68 61 261 2 1 0 404

Closed 21 10 366 267 923 11 0 6 1604

Museum (302, total)

Open 2 7 20 7 6 0 0 0 42

Closed 33 20 168 16 22 0 1 0 260

B.4b

Percentage of () tokens according to syllable type and age

Young

Open 0.00 0.00 2.24 5.13 90.38 1.60 0.64 0.00

Closed 0.00 0.00 2.81 10.52 84.47 0.60 0.20 1.40

Old

Open 0.74 1.98 16.83 15.10 64.60 0.50 0.25 0.00

Closed 1.31 0.62 22.82 16.65 57.54 0.69 0.00 0.37

Museum

Open 4.76 16.67 47.62 16.67 14.29 0.00 0.00 0.00

Closed 12.69 7.69 64.62 6.15 8.46 0.00 0.38 0.00

344

B.5a


Total

Stop 18 17 103 106 555 7 0 0 806

Fricative 12 9 120 74 342 4 0 0 561

Nasal 6 1 98 59 355 2 1 0 522

Approximant 22 13 327 223 1021 10 5 20 1641

Affricate 0 0 2 0 3 0 0 0 5

None 1 5 7 10 61 1 0 0 85

B.5b


Stop 2.23 2.11 12.78 13.15 68.86 0.87 0.00 0.00

Fricative 2.14 1.60 21.39 13.19 60.96 0.71 0.00 0.00

Nasal 1.15 0.19 18.77 11.30 68.01 0.38 0.19 0.00

Approximant 1.34 0.79 19.93 13.59 62.22 0.61 0.30 1.22

Affricate 0.00 0.00 40.00 0.00 60.00 0.00 0.00 0.00

None 1.18 5.88 8.24 11.76 71.76 1.18 0.00 0.00

B.6a


Total

Labial 21 9 200 177 595 5 2 20 1029

Coronal 36 31 448 275 1538 18 4 0 2350

Dorsal 1 0 1 9 140 0 0 0 151

Glottal 0 0 1 1 3 0 0 0 5

None 1 5 7 10 61 1 0 0 85

B.6b


Labial 2.04 0.87 19.44 17.20 57.82 0.49 0.19 1.94

Coronal 1.53 1.32 19.06 11.70 65.45 0.77 0.17 0.00

Dorsal 0.66 0.00 0.66 5.96 92.72 0.00 0.00 0.00

Glottal 0.00 0.00 20.00 20.00 60.00 0.00 0.00 0.00

None 1.18 5.88 8.24 11.76 71.76 1.18 0.00 0.00

345

B.7a


Total

Voiced 37 28 288 228 1138 17 5 0 1741

Voiceless 19 12 327 210 990 7 1 20 1586

None 3 5 42 34 209 0 0 0 293

B.7b


Voiced 2.13 1.61 16.54 13.10 65.36 0.98 0.29 0.00

Voiceless 1.20 0.76 20.62 13.24 62.42 0.44 0.06 1.26

None 1.02 1.71 14.33 11.60 71.33 0.00 0.00 0.00

B.8a

Number of () tokens according to following voice by Museum speakers

Total

Voiced 17 17 99 19 18 0 1 0 171

Voiceless 17 8 80 2 8 0 0 0 115

None 1 2 9 2 2 0 0 0 16

B.8b

Percentage of () variants according to following voice by Museum speakers

Voiced 9.94 9.94 57.89 11.11 10.53 0.00 0.58 0.00

Voiceless 14.78 6.96 69.57 1.74 6.96 0.00 0.00 0.00

None 6.25 12.50 56.25 12.50 12.50 0.00 0.00 0.00

346

B.9a

Number of () tokens according to following voice by Older speakers

Total

Voiced 20 11 180 154 596 8 1 0 970

Voiceless 2 4 225 149 508 5 0 6 899

None 2 3 29 25 80 0 0 0 139

B.9b

Percentage of () variants according to following voice by Older speakers

Voiced 2.06 1.13 18.56 15.88 61.44 0.82 0.10 0.00

Voiceless 0.22 0.44 25.03 16.57 56.51 0.56 0.00 0.67

None 1.44 2.16 20.86 17.99 57.55 0.00 0.00 0.00

B.10a

Number of () tokens according to following voice by Younger speakers

Total

Voiced 0 0 9 55 524 9 3 0 600

Voiceless 0 0 22 59 474 2 1 14 572

None 0 0 4 7 127 0 0 0 138

B.10b

Percentage of () variants according to following voice by Younger speakers

Voiced 0.00 0.00 1.50 9.17 87.33 1.50 0.50 0.00

Voiceless 0.00 0.00 3.85 10.31 82.87 0.35 0.17 2.45

None 0.00 0.00 2.90 5.07 92.03 0.00 0.00 0.00

347

B.11a


Total

Stop 22 17 359 230 957 7 2 20 1614

Fricative 18 10 128 89 447 3 0 0 695

Nasal 15 13 115 106 627 9 2 0 887

Approximant 0 0 5 6 31 2 0 0 44

Flap 0 0 5 1 3 0 0 0 9

Affricate 0 0 0 0 1 0 0 0 1

Vowel 1 0 3 6 62 3 2 0 77

None 3 5 42 34 209 0 0 0 293

B.11b

Overall percentages of () variants according to following manner of articulation

Stop 1.4 1.1 22.2 14.3 59.3 0.4 0.1 1.2

Fricative 2.6 1.4 18.4 12.8 64.3 0.4 0.0 0.0

Nasal 1.7 1.5 13.0 12.0 70.7 1.0 0.2 0.0

Approximant 0.0 0.0 11.4 13.6 70.5 4.5 0.0 0.0

Flap 0.0 0.0 55.6 11.1 33.3 0.0 0.0 0.0

Affricate 0.0 0.0 0.0 0.0 100 0.0 0.0 0.0

Vowel 1.3 0.0 3.9 7.8 80.5 3.9 2.6 0.0

None 1.0 1.7 14.3 11.6 71.3 0.0 0.0 0.0

348

B.12a


Total

Labial 22 19 131 93 498 7 0 0 770

Coronal 31 14 288 226 1042 14 4 0 1619

Dorsal 0 1 9 18 51 0 0 0 79

Glottal 2 6 184 95 475 0 0 20 782

Vowel 1 0 3 6 62 3 2 0 77

None 3 5 42 34 209 0 0 0 293

B.12b


Labial 2.9 2.5 17.0 12.1 64.7 0.9 0.0 0.0

Coronal 1.9 0.9 17.8 14.0 64.4 0.9 0.2 0.0

Dorsal 0.0 1.3 11.4 22.8 64.6 0.0 0.0 0.0

Glottal 0.3 0.8 23.5 12.1 60.7 0.0 0.0 2.6

Vowel 1.3 0.0 3.9 7.8 80.5 3.9 2.6 0.0

None 1.0 1.7 14.3 11.6 71.3 0.0 0.0 0.0

B.13a

Overall number of () tokens preceding /l/

Total

Coda L 0 1 3 20 10 62 96

Non Coda L 2 0 0 21 19 80 122

B.13b

Overall percentage of () variants preceding /l/

Coda L 0.00 1.04 3.13 20.83 10.42 64.58

Non Coda L 1.64 0.00 0.00 17.21 15.57 65.57

349

B.14a

Overall comparison of the number of () tokens according to /l/ type

Total

Coda L 0 1 3 20 10 62 96

Non Coda L 1 0 0 9 5 0 15

Island 1 0 0 12 14 80 107

B.14b

Overall percentage comparison of () variants according to /l/ type

Coda L 0.00 1.04 3.13 20.83 10.42 64.58

Non Coda L 6.67 0.00 0.00 60.00 33.33 0.00

Island 0.93 0.00 0.00 11.21 13.08 74.77

B.15a

Overall number of diphthongal and monophthongal tokens for the token ‘Island’

Diphthong Monophthong Total

Island 27 80 107

B.15b

Overall percentage of diphthongal and monophthongal variants for the token ‘Island’

Diphthong Monophthong

Island 25.2 74.8

B.16a

Number of () tokens for ‘Island’ by Museum speakers


Island 2 7 9

B.16b

Percentage of () variants for ‘Island’ by Museum speaker


Island 22.2 77.8

350

B.17a

Number of () tokens for ‘Island’ by Older speakers


Island 20 69 89

B.17b

Percentage of () variants for ‘Island’ by Older speakers


Island 22.5 77.5

B.18a

Number of () tokens for ‘Island’ by Younger speakers


Island 5 4 9

B.18b

Percentage of () variants for ‘Island’ by Younger speakers


Island 55.6 44.4

B.20a

An overall comparison of () tokens according to the functions of LIKE

Total

Grammatical Functions 45 41 211 1 0 298

Discursive Functions 19 60 368 0 1 448

B.20b

The overall percentages of () variants according to the functions of LIKE

Grammatical Functions 15.10 13.76 70.81 0.34 0.00

Discursive Functions 4.24 13.39 82.14 0.00 0.22

351

B.21a

Number of () tokens according to the function of LIKE by Older speakers

Total



B.21b

Percentage of () variants according to the functions of LIKE by Older speakers



B.22a

Number of () tokens according to the function of LIKE by Younger speakers

Total



B.22b

Percentage of () variants according to the function of LIKE by Younger speakers



352

From Preceding Phonological Environment Chapter 6.3.2

Museum

B.23a

Number of () tokens by Museum speakers according to preceding manner of

articulation

Total

Stop 5 11 37 5 9 0 0 0 67

Fricative 5 5 33 6 4 0 0 0 53

Nasal 5 0 37 4 3 0 0 0 49

Approximant 20 9 81 6 11 0 1 0 128

Affricate 0 0 0 0 0 0 0 0 0

None 0 2 0 2 1 0 0 0 5

B.23b

Percentage of () variants by Museum speakers according to preceding manner of

articulation

Stop 7.46 16.42 55.22 7.46 13.43 0.00 0.00 0.00

Fricative 9.43 9.43 62.26 11.32 7.55 0.00 0.00 0.00

Nasal 10.20 0.00 75.51 8.16 6.12 0.00 0.00 0.00

Approximant 15.63 7.03 63.28 4.69 8.59 0.00 0.78 0.00

Affricate 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

None 0.00 40.00 0.00 40.00 20.00 0.00 0.00 0.00

353

Older

B.24a

Number of () tokens by Older speakers according to preceding manner of articulation

Total

Stop 13 6 64 77 286 5 0 0 451

Fricative 7 4 84 48 165 2 0 0 310

Nasal 1 1 51 41 207 2 0 0 303

Approximant 2 4 226 155 495 4 1 6 893

Affricate 0 0 2 0 1 0 0 0 3

None 1 3 7 7 30 0 0 0 48

B.24b

Percentage of () variants by Older speakers according to preceding manner of

articulation

Stop 2.88 1.33 14.19 17.07 63.41 1.11 0.00 0.00

Fricative 2.26 1.29 27.10 15.48 53.23 0.65 0.00 0.00

Nasal 0.33 0.33 16.83 13.53 68.32 0.66 0.00 0.00

Approximant 0.22 0.45 25.31 17.36 55.43 0.45 0.11 0.67

Affricate 0.00 0.00 66.67 0.00 33.33 0.00 0.00 0.00

None 2.08 6.25 14.58 14.58 62.50 0.00 0.00 0.00

354

Younger

B.25a

Number of () tokens by Younger speakers according to preceding manner of

articulation

Total

Stop 0 0 2 24 260 2 0 0 288

Fricative 0 0 3 20 173 2 0 0 198

Nasal 0 0 10 14 145 0 1 0 170

Approximant 0 0 20 62 515 6 3 14 620

Affricate 0 0 0 0 2 0 0 0 2

None 0 0 0 1 30 1 0 0 32

B.25b

Percentage of () variants by Younger speakers according to preceding manner of

articulation

Stop 0.00 0.00 0.69 8.33 90.28 0.69 0.00 0.00

Fricative 0.00 0.00 1.52 10.10 87.37 1.01 0.00 0.00

Nasal 0.00 0.00 5.88 8.24 85.29 0.00 0.59 0.00

Approximant 0.00 0.00 3.23 10.00 83.06 0.97 0.48 2.26

Affricate 0.00 0.00 0.00 0.00 100.00 0.00 0.00 0.00

None 0.00 0.00 0.00 3.13 93.75 3.13 0.00 0.00

355

Museum

B.26a

Number of () tokens by Museum speakers according to preceding place of articulation

Total

Labial 13 6 42 8 3 0 0 0 72

Coronal 22 19 146 13 23 0 1 0 224

Dorsal 0 0 0 0 1 0 0 0 1

Glottal 0 0 0 0 0 0 0 0 0

None 0 2 0 2 1 0 0 0 5

B.26b

Percentage of () variants by Museum speakers according to preceding place of

articulation

Labial 18.06 8.33 58.33 11.11 4.17 0.00 0.00 0.00

Coronal 9.82 8.48 65.18 5.80 10.27 0.00 0.45 0.00

Dorsal 0.00 0.00 0.00 0.00 100.00 0.00 0.00 0.00

Glottal 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

None 0.00 40.00 0.00 40.00 20.00 0.00 0.00 0.00

356

Older

B.27a

Number of () tokens by Older speakers according to preceding place of articulation

Total

Labial 8 3 132 105 262 3 1 6 520

Coronal 14 12 294 212 872 10 0 0 1414

Dorsal 1 0 0 3 17 0 0 0 21

Glottal 0 0 1 1 3 0 0 0 5

None 1 3 7 7 30 0 0 0 48

B.28b

Percentage of () variants by Older speakers according to preceding place of

articulation

Labial 1.54 0.58 25.38 20.19 50.38 0.58 0.19 1.15

Coronal 0.99 0.85 20.79 14.99 61.67 0.71 0.00 0.00

Dorsal 4.76 0.00 0.00 14.29 80.95 0.00 0.00 0.00

Glottal 0.00 0.00 20.00 20.00 60.00 0.00 0.00 0.00

None 2.08 6.25 14.58 14.58 62.50 0.00 0.00 0.00

357

Younger

B.28a

Number of () tokens by Younger speakers according to preceding place of articulation

Total

Labial 0 0 26 64 330 2 1 14 437

Coronal 0 0 8 50 643 8 3 0 712

Dorsal 0 0 1 6 122 0 0 0 129

Glottal 0 0 0 0 0 0 0 0 0

None 0 0 0 1 30 1 0 0 32

B.28b

Percentage of () variants by Younger speakers according to preceding place of

articulation

Labial 0.00 0.00 5.95 14.65 75.51 0.46 0.23 3.20

Coronal 0.00 0.00 1.12 7.02 90.31 1.12 0.42 0.00

Dorsal 0.00 0.00 0.78 4.65 94.57 0.00 0.00 0.00

Glottal 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

None 0.00 0.00 0.00 3.13 93.75 3.13 0.00 0.00

358

Appendix C

359

Appendix C


of the variable () (see Chapter 8)

C.1


Total

Young 0 4 80 0 84

Old 44 23 103 5 175

Museum 8 1 18 0 27

Total 52 28 201 5 286

C.2a


Total

Young

Male 0 2 43 0 45

Female 0 2 37 0 39

Old

Male 15 7 62 5 89

Female 29 16 41 0 86

Museum

Male 8 1 18 0 27

C.2b


Young

Male 0.00 4.44 95.56 0.00

Female 0.00 5.13 94.87 0.00

Old

Male 16.85 7.87 69.66 5.62

Female 33.72 18.60 47.67 0.00

Museum

Male 29.63 3.70 66.67 0.00

360

C.3a

Overall number of () tokens according to syllable-type

Total

Open 23 13 108 0 144

Closed 29 15 93 5 142

Total 52 28 201 5 286

C.3b

Overall percentage of () variants according to syllable-type

Open 15.97 9.03 75.00 0.00

Closed 20.42 10.56 65.49 3.52

Total 18.18 9.79 70.28 1.75

C.4a

Number of () tokens produced by Males according to age and syllable type

Males TotalYoung (45, total)

Open 2 21 23

Closed 22 22

Old (89, total)

Open 7 3 36 46

Closed 8 4 26 5 43

Museum (27, total)

Open 5 17 22

Closed 3 1 1 5

C.4b

Percentage of () variants produced by Males according to age and syllable type

Young

Open 0.00 8.70 91.30 0.00

Closed 0.00 0.00 100 0.00

Old

Open 15.22 6.52 78.26 0.00

Closed 18.60 9.30 60.47 11.63

Museum

Open 22.73 0.00 77.27 0.00

Closed 60.00 20.00 20.00 0.00

361

C.5a

Number of () tokens produced by females according to age and syllable type

Females TotalYoung (39, total)

Open 1 12 13

Closed 1 25 26

Old (86, total)

Open 11 7 22 40

Closed 18 9 19 46

C.5b

Percentage of () variants produced by females according to age and syllable type

Young

Open 0.00 7.69 92.31 0.00

Closed 0.00 3.85 96.15 0.00

Old

Open 27.50 17.50 55.00 0.00

Closed 39.13 19.57 41.30 0.00

C.6a


TotalLabial 45 18 94 157

Coronal 5 6 66 77

Dorsal 1 2 3

Word-initial 2 4 37 3 46

/sp/ 3 3

C.6b


Labial 28.66 11.46 59.87 0.00

Coronal 6.49 7.79 85.71 0.00

Dorsal 0.00 0.00 33.33 66.67

Word-initial 4.35 8.70 80.43 6.52

/sp/ 0.00 0.00 100 0.00

362

C.7a

Number of () tokens produced by Museum speakers according to preceding place of

articulation

TotalLabial 7 3 10

Coronal 1 8 9

Word-initial 1 7 8

C.7b

Percentage of () variants produced by Museum speakers according to preceding place

of articulation

Labial 70.00 0.00 30.00 0.00

Coronal 0.00 11.11 88.89 0.00

Word-initial 12.50 0.00 87.50 0.00

C.8a

Number of () tokens produced by Older speakers according to preceding place of

articulation

TotalLabial 38 14 41 93

Coronal 5 5 33 43

Dorsal 2 2

Word-initial 1 4 26 3 34

/sp/ 3 3

C.8b

Percentage of () variants produced by Older speakers according to preceding place of

articulation

Labial 40.86 15.05 44.09 0.00

Coronal 11.63 11.63 76.74 0.00

Dorsal 0.00 0.00 0.00 100

Word-initial 2.94 11.76 76.47 8.82

/sp/ 0.00 0.00 100 0.00

363

C.9a

Number of () tokens produced by Younger speakers according to preceding place of

articulation

TotalLabial 4 50 54

Coronal 25 25

Dorsal 1 1

Word-initial 4 4

C.9b

Percentage of () variants produced by Younger speakers according to preceding place

of articulation

Labial 0.00 7.41 92.59 0.00

Coronal 0.00 0.00 100 0.00

Dorsal 0.00 0.00 100 0.00

Word-initial 0.00 0.00 100 0.00

C.10a


TotalStop 43 19 85 2 149

Fricative 2 15 17

Nasal 4 2 14 20

Affricate 1 45 46

Approximant 1 2 2 5

/sp/ 3 3

None 2 4 37 3 46

C.10b


Stop 28.86 12.75 57.05 1.34

Fricative 11.76 0.00 88.24 0.00

Nasal 20.00 10.00 70.00 0.00

Affricate 0.00 2.17 97.83 0.00

Approximant 20.00 40.00 40.00 0.00

/sp/ 0.00 0.00 100 0.00

None 4.35 8.70 80.43 6.52

364

C.11a

Number of lexically derived BOY tokens produced by Museum speakers

Totalboy 4 0 2 0 6

boys 3 0 0 0 3

Total 7 0 2 0 9

C.11b

Percentage of lexically derived BOY tokens produced by Museum speakers

boy 66.67 0.00 33.33 0.00

boys 100 0.00 0.00 0.00

C.12a

Number of lexically derived BOY tokens produced by Older speakers

Totalboy 16 5 14 0 35

boys 18 7 8 0 33

boy's 0 1 0 0 1

Total 34 13 22 0 69

C.12b

Percentage of lexically derived BOY tokens produced by Older speakers

boy 45.71 14.29 40.00 0.00

boys 54.55 21.21 24.24 0.00

boy's 0.00 100 0.00 0.00

365

C.13a

Number of lexically derived BOY tokens produced by Younger speakers

Totalboy 0 3 7 0 10

boys 0 1 2 0 3

boyfriend 0 0 1 0 1

boyish 0 0 1 0 1

Total 0 4 11 0 15

C.13b

Percentage of lexically derived BOY tokens produced by Younger speakers

boy 0.00 30.00 70.00 0.00

boys 0.00 33.33 66.67 0.00

boyfriend 0.00 0.00 100 0.00

boyish 0.00 0.00 100 0.00

C.14a

Number of BOY tokens compared to that of ‘other’ tokens by Museum speakers

TotalBOY 7 0 2 0 9

other 1 1 16 0 18

Total 8 1 18 0 27

C.14b

Percentage of BOY variants compared to that of ‘other’ tokens by Museum speakers

BOY 77.78 0.00 22.22 0.00

other 5.56 5.56 88.89 0.00

366

C.15a

Number of BOY tokens compared to that of ‘other’ tokens by Older speakers

TotalBOY 34 13 22 0 69

other 10 10 81 5 106

Total 44 23 103 5 175

C.15b

Percentage of BOY variants compared to that of ‘other’ tokens by Older speakers

BOY 49.28 18.84 31.88 0.00

other 9.43 9.43 76.42 4.72

367

Following Phonological Environment (which was not included in the overall analysis).

From Chapter 8.3.2

C.16a


TotalLabial 2 2

Coronal 32 20 161 5 218

Glottal 1 1

Word-final 20 8 37 65

C.16b


Labial 0.00 0.00 100.00 0.00

Coronal 14.68 9.17 73.85 2.29

Glottal 0.00 0.00 100.00 0.00

Word-final 30.77 12.31 56.92 0.00

C.16c

The Overall Percentage for () Variants by Mersea Island Speakers According to

Following Place of Articulation

0

20

40

60

80

100

Labial Coronal Glottal Word-final

368

C.17a

Overall number of () tokens produced by Museum speakers according to following

place of articulation

Total

Coronal 4 1 15 20

Word-final 4 3 7

C.17b

Percentage of () variants produced by Museum speakers according to following place

of articulation

Coronal 20.00 5.00 75.00 0.00

Word-final 57.14 0.00 42.86 0.00

C.17c

Percentage of () Variants by Museum Speakers According to Following Place of

Articulation

0

20

40

60

80

100

Coronal Word-final

369

C.18a

Overall number of () tokens produced by Old speakers according to following place of

articulation

TotalLabial 1 1

Coronal 28 18 75 5 126

Glottal 1 1

Word-final 16 5 26 47

C.18.b

Percentage of () variants produced by Old speakers according to following place of

articulation

Labial 0.00 0.00 100.00 0.00

Coronal 22.22 14.29 59.52 3.97

Glottal 0.00 0.00 100.00 0.00

Word-final 34.04 10.64 55.32 0.00

C.18c

Percentage of () Variants by Old Speakers According to Following Place of

Articulation

0

20

40

60

80

100

Labial Coronal Glottal Word-final

370

C.19a

Overall number of () tokens produced by Younger speakers according to following

place of articulation

TotalLabial 1 1

Coronal 1 71 72

Word-final 3 8 11

C.19b

Percentage of () variants produced by Younger speakers according to following place

of articulation

Labial 0.00 0.00 100.00 0.00

Coronal 0.00 1.39 98.61 0.00

Word-final 0.00 27.27 72.73 0.00

C.19c

Percentage of () Variants by Younger Speakers According to Following Place of

Articulation

0

20

40

60

80

100

Labial Coronal Word-final

371

C.20a


TotalStop 1 2 7 10


Nasal 1 1 69 71


None 20 8 37 65

C.20b

Overall percentage of () variants according to following manner of articulation

Stop 10.00 20.00 70.00 0.00

Fricative 26.61 13.76 59.63 0.00

Nasal 1.41 1.41 97.18 0.00

Approximant 3.23 6.45 74.19 16.13

None 30.77 12.31 56.92 0.00

C.20c


Following Manner of Articulation

0

20

40

60

80

100

Stop Fricative Nasal Approximant None

372

C.21a



TotalFricative 4 7 11

Nasal 1 8 9

None 4 3 7

C.21b

Percentage of () variants produced by Museum speakers according to following


Fricative 36.36 0.00 63.64 0.00

Nasal 0.00 11.11 88.89 0.00

None 57.14 0.00 42.86 0.00

C.21c

Percentage of () Variants by Museum Speakers According to Following Manner of

Articulation

0

20

40

60

80

100

Fricative Nasal None

373

C.22a

Overall number of () tokens produced by Older speakers according to following


TotalStop 1 2 3 6


Nasal 1 29 30


None 16 5 26 47

C.22b

Percentage of () variants produced by Older speakers according to following manner

of articulation

Stop 16.67 33.33 50.00 0.00

Fricative 34.25 19.18 46.58 0.00

Nasal 3.33 0.00 96.67 0.00

Approximant 5.26 10.53 57.89 26.32

None 34.04 10.64 55.32 0.00

C.22c

Percentage of () Variants by Older Speakers According to Following Manner of

Articulation

0.00

20.00

40.00

60.00

80.00

100.00


374

C.23a



TotalStop 4 4

Fricative 1 24 25

Nasal 32 32

Approximant 12 12

None 3 8 11

C.23b

Percentage of () variants produced by Younger speakers according to following


Stop 0.00 0.00 100.00 0.00

Fricative 0.00 4.00 96.00 0.00

Nasal 0.00 0.00 100.00 0.00

Approximant 0.00 0.00 100.00 0.00

None 0.00 27.27 72.73 0.00

C.23c

Percentage of () Variants by Younger Speakers According to Following Manner of

Articulation

0

20

40

60

80

100


375

C.24a


TotalVoiced 28 17 110 5 160

Voiceless 4 3 54 61

None 20 8 37 65

C.24b


Voiced 17.50 10.63 68.75 3.13

Voiceless 6.56 4.92 88.52 0.00

None 30.77 12.31 56.92 0.00

C.24c


Following Voice

0

20

40

60

80

100


376

C.25a


voice

TotalVoiced 3 1 8 12

Voiceless 1 7 8

None 4 3 7

C.25b

Percentage of () variants produced by Museum speakers according to following voice

Voiced 25.00 8.33 66.67 0.00

Voiceless 12.50 0.00 87.50 0.00

None 57.14 0.00 42.86 0.00

C.25c

Percentage of () Variants by Museum Speakers According to Following Voice

0

20

40

60

80

100


377

C.26a

Overall number of () tokens produced by Older speakers according to following voice

TotalVoiced 25 15 51 5 96

Voiceless 3 3 26 32

None 16 5 26 47

C.26b

Percentage of () variants produced by Older speakers according to following voice

Voiced 26.04 15.63 53.13 5.21

Voiceless 9.38 9.38 81.25 0.00

None 34.04 10.64 55.32 0.00

C.26c

Percentage of () Variants by Older Speakers According to Following Voice

0

20

40

60

80

100


378

C.27a


voice

TotalVoiced 1 51 52

Voiceless 21 21

None 3 8 11

C.27b

Percentage of () variants produced by Younger speakers according to following voice

Voiced 0.00 1.92 98.08 0.00

Voiceless 0.00 0.00 100.00 0.00

None 0.00 27.27 72.73 0.00

C.27c

Percentage of () Variants by Younger Speakers According to Following Voice

0

20

40

60

80

100


379

Preceding Manner of Articulation, from Chapter 8.3.2.2

C.28a

Number of () tokens according to following manner of articulation by Museum

speakers

Total

Stop 7 3 10

Affricate 1 8 9

None 1 7 8

C.28b

Percentage of () tokens according to following manner of articulation by Museum

speakers

Stop 70.00 0.00 30.00 0.00

Affricate 0.00 11.11 88.89 0.00

None 12.50 0.00 87.50 0.00

380

C.29a

Number of () tokens according to following manner of articulation by Older speakers

Total

Stop 36 15 44 2 97

Fricative 2 1 3

Nasal 4 2 2 8

Affricate 26 26

Approximant 1 2 1 4

/sp/ 3 3

None 1 4 26 3 34

C.29b

Percentage of () tokens according to following manner of articulation by Older

speakers

Stop 37.11 15.46 45.36 2.06

Fricative 66.67 0.00 33.33 0.00

Nasal 50.00 25.00 25.00 0.00

Affricate 0.00 0.00 100.00 0.00

Approximant 25.00 50.00 25.00 0.00

[sp] 0.00 0.00 100.00 0.00

None 2.94 11.76 76.47 8.82

381

C.30a

Number of () tokens according to following manner of articulation by Younger

speakers

Total

Stop 4 38 42

Fricative 14 14

Nasal 12 12

Affricate 11 11

Approximant 1 1

None 4 4

C.30b

Percentage of () tokens according to following manner of articulation by Younger

speakers

Stop 0.00 9.52 90.48 0.00

Fricative 0.00 0.00 100.00 0.00

Nasal 0.00 0.00 100.00 0.00

Affricate 0.00 0.00 100.00 0.00

Approximant 0.00 0.00 100.00 0.00

None 0.00 0.00 100.00 0.00

382

Appendix D

383

Appendix D

In this appendix, you will find the examples from the Mersea data for each of the

variables examined in this thesis with respect to preceding and following environments.

1. The variable (a)

a) Preceding Phonological Environment

Preceding Consonant Examples

[p] pound, pounding, pounce, pouncing

[b] about, bounce, bound, boundary

[m] amount, mouth, mouse, mountains

[w] wound-up

[f] foundation, found, fowl

[v] voucher

[t] town, towns

[d] down, doubt, endowment

[n] now, announce, pronounce

[] thousand, without

[s] sound, sounded, south, Southend,

[] shout, shouted, shouting

[l] loud, lounge, allow, plough

[r] round, around, ground, brown

[k] cows, count, council, accounting

[] dug-outs

[h] house, house how, hound

[sk] scout, scouts

Word Initial out, outings, ounce

384

1. The variable (a)

b) Following Phonological Environment

Following Consonant Examples

[m] ploughman, endowment

[w] allowing

[f] south (with final th-fronting)

[t] out, about, shout, without

[d] cloud, loud, ploughed, allowed

[n] bound, down, fount, crown

[] out, about (/t/ to flap due to connected speech)

[] mouth, south

[s] house, mouse, houseboat, greenhouse

[z] houses, thousand, ploughs, rows

[l] towel, fowl

[] voucher, Crouch (Street)

[] lounge

[h] cowhide

[] out, about, outside, output

[] Haward (family name), allowance, now-a-days

Word Final how, now, bow, allow

385

2. The variable (a)



[p] pipe, pies, pine, pilot

[b] bike, bite, bible, combined

[m] mile, mine, reminds, minus

[w] white, while, why, quite, choir

[f] find, fight, five, fine

[v] advice, vice, survive, televised

[t] tie, time, advertise, sometime

[d] dive, die, dining, dicing

[n] knife, night, nice, modernise

[] thigh, empathise

[s] sight, size, aside, excited

[z] design

[] shy, shiny, shining

[l] like, life, lice, realise, apply

[r] rival, right, alright, try

[] child

[] fragile, apologise, energised, gigantic

[k] kind, kite, archive

[] guy, guide, gynaecologist, guidelines

[h] high, hiding, hive, hind

[] sometimes (with /t/ glottalisation)

[sp] spy, spider, spicy

[sm] Smile

[st] style, sty

[sl] slide, slice, slight, slightest

[sn] Snipe

[sk] sky, skylark

Word Initial eye, ice, idea, Island

386

2. The variable (a)



[p] ipod, pipes, hyperlinks, type

[b] tribe, bible, library, bribery

[m] time, crime, miming, rhyme

[w] highway, plywood

[f] life, wife, lifeboat, decipher

[v] five, rival, driving, private

[t] right, site, might, exciting

[d] tried, died, tide, qualified

[n] find, kind, China, nine

[] might, quite, right (/t/ to flap in connected speech)

[] Hythe, either

[s] nice, twice, bicycle, price

[z] guys, otherwise, exercise, dies

[l] mile, while, child, file

[r] Irish, spiral, siren,

[j] trying, flying, highest, Ryan

[] hygiene

[k] like, biker, psychology, cycle

[] quite, right, delight, fight

Vowel triangle, diagonal, Diane, highest, quiet

Word Final tie, high, why, guy

387

3. The variable ()



[p] point, poisonous, appointment

[b] boy, boiler, boiled

[v] voice, avoid

[t] toilet, toying, toys

[n] noise, annoying

[l] alloys, employed

[r] destroy

[] choice

[] join, enjoy, joint

[k] coy, coil

[sp] spoil

Word Initial oyster, oil

3. The variable ()



[m] enjoyment

[f] boyfriend

[d] avoid, enjoyed, paranoid, employed

[n] join, point, joined, disappoint

[s] voice, choice, oysters

[z] boys, noise, alloys

[l] oil, toilet, boiled, coil

[j] annoying, toying, boyish

[] pointing

Vowel

Word Final boy, joy, destroy

388

Appendix E

389

Appendix E

In this appendix, you will find the (stressed) vowel inventories of two speakers from the

Mersea corpus used in this research.

1) Speaker BC: Museum Male, b. 1883

Wells’ Lexical Set

Reference

Realisation Wells’ Lexical Set

Reference

Realisation

KIT THOUGHT :

DRESS ~ GOAT ~ , ~

1 , 2

, ~3

TRAP : ~ : ~

LOT : ~ GOOSE :

STRUT ~ PRICE

FOOT CHOICE

BATH : MOUTH

CLOTH : ~ ~ : NEAR

NURSE : SQUARE

FLEECE i: ~ i: START :

FACE ~ NORTH :

PALM : FORCE :

CURE :4

Notes:

1. found in go and no

2. found in most

3. found in non-main stressed syllable of, for example, meadow

4. vowel quality based on poor and your

Additional notes:

5. r-coloured vowels, particularly in the NURSE set

390

2) Speaker SC: Young Female, b 1989

Wells’ Lexical Set

Reference

Realisation Wells’ Lexical Set

Reference

Realisation

KIT THOUGHT :

DRESS GOAT

TRAP : GOOSE :

LOT PRICE

STRUT ~ CHOICE

FOOT ~ ~ MOUTH

BATH : NEAR

CLOTH SQUARE :

NURSE : START :

FLEECE i: NORTH :

FACE FORCE :

PALM : CURE :1

Notes:

1. vowel based on your and sure

Documents

A Sociophonological Analysis of Mersea Island Englishroa.rutgers.edu/content/article/files/1208_amos_1.pdf · IV ABSTRACT This thesis presents a socio-phonological analysis of three